Using trait and phylogenetic diversity to evaluate the generality of
the stress-dominance hypothesis in eastern North American tree
Jessica R. Coyle, Fletcher W. Halliday, Bianca E. Lopez, Kyle A. Palmquist, Peter A. Wilfahrt
and Allen H. Hurlbert
J. R. Coyle (firstname.lastname@example.org), F. W. Halliday and A. H. Hurlbert, Dept of Biology, Univ. of North Carolina at Chapel Hill, Chapel Hill,
NC 27599-3280, USA. – B. E. Lopez, K. A. Palmquist, P. A. Wilfahrt and AHH, Curriculum for the Environment and Ecology, Univ. of
North Carolina at Chapel Hill, Chapel Hill, NC 27599-3275, USA.
e stress-dominance hypothesis (SDH) is a model of community assembly predicting that the relative importance of
environmental ﬁltering increases and competition decreases along a gradient of increasing environmental stress. Tests
of the SDH at limited spatial scales have thus far demonstrated equivocal support and no prior study has assessed the
generality of the SDH at continental scales. We examined over 53 000 tree communities spanning the eastern United
States to determine whether functional trait variation and phylogenetic diversity support the SDH for gradients of
water and soil nutrient availability. is analysis incorporated two complementary datasets, those of the U.S. Forest
Service Forest Inventory and Analysis National program and the Carolina Vegetation Survey, and was based on three
ecologically important traits: leaf nitrogen, seed mass, and wood density. We found that mean trait values were weakly
correlated with water and soil nutrient availability, but that trait diversity did not vary consistently along either gradient.
is did not conform to trait variation expected under the SDH and instead suggested that environmental ﬁlters
structure tree communities throughout both gradients, without evidence for an increased role of competition in less
stressful environments. Phylogenetic diversity of communities was principally driven by the ratio of angiosperms to
gymnosperms and therefore did not exhibit the pattern of variation along stress gradients expected under the SDH. We
conclude that the SDH is not a general paradigm for all eastern North American tree communities, although it may
operate in certain contexts.
Ecological communities are expected to be structured by a
variety of stochastic and deterministic processes. Two deter-
ministic processes thought to play a strong role in determin-
ing the coexistence of species in the same trophic level are
interspeciﬁc competition and environmental ﬁltering, where
species are excluded from a community due to an inability to
survive and reproduce in a given physical environment.
General rules for predicting the relative importance of these
processes in diﬀerent contexts are still largely unresolved
(HilleRisLambers et al. 2012). One compelling community
assembly model predicts that environmental ﬁltering will be
more important in structuring communities in stressful
environments, while competitive interactions will be more
important in benign environments (depicted in Weiher and
Keddy 1995). is hypothesis, which we refer to as the
stress-dominance hypothesis (terminology adapted from
Swenson and Enquist 2007), derives from the expectation
that the importance of competition in plant communities
declines with increasing environmental stress (Grime 1977)
and is consistent with modern theory predicting that ﬁtness
diﬀerences change along abiotic gradients (HilleRisLambers
et al. 2012).
Testing the generality of the stress-dominance hypothesis
(SDH) is diﬃcult. e experiments required to identify
competition and environmental ﬁltering are infeasible in
many types of communities (for example, among long-lived
organisms) and rarely cover a suﬃciently broad geographic
extent to conﬁrm that the hypothesis is generalizable. Large-
scale observational studies provide a unique opportunity to
evaluate the generality of the SDH because 1) the number of
communities that can be analyzed may be several orders of
magnitude larger than in traditional ﬁeld studies, 2) large-
scale studies tend to span broader environmental gradients
related to the processes of interest, and 3) the occurrence of
strong relationships amidst the ecological heterogeneity of a
large observational data set provides more persuasive evi-
dence of generality. Nevertheless, parsing ecological pro-
cesses in a non-experimental setting is notoriously diﬃcult,
especially when making inferences from patterns of species
diversity and community composition (Gotelli and Graves
Ecography 37: 814–826, 2014
© 2014 e Authors. Ecography © 2014. Nordic Society Oikos
Subject Editor: Dominique Gravel. Accepted 12 January 2014
1996). Ecologists are increasingly using information about
species traits and phylogenetic relationships to strengthen
observational pattern-based inference of the mechanisms of
community assembly (Webb et al. 2002, Emerson and
Gillespie 2008, Kraft et al. 2008, Cavender-Bares et al. 2009,
Spasojevic and Suding 2012). e variation of certain traits
within a community can indicate the relative strength of
processes such as competition and environmental ﬁltering
since traits diﬀerentially mediate an organism’s ability to use
and obtain resources and to tolerate environmental stressors
(Keddy 1992, Tilman 2004). Where trait information has
been unavailable, the phylogenetic structure of communi-
ties has frequently been used as a proxy for functional struc-
ture (Bryant et al. 2008, Cadotte et al. 2008, Graham and
Functional and phylogenetic diversity have recently
been used to elucidate processes structuring a wide range of
communities (Ricotta and Moretti 2011), and several stud-
ies have found patterns consistent with the SDH (Swenson
and Enquist 2007, Kluge and Kessler 2011, Machac et al.
2011, Graham et al. 2012, Mason et al. 2012, Spasojevic
and Suding 2012). However, most studies encompass a
relatively small geographic extent and focus on a single
community or habitat type. While this allows researchers
to collect detailed data on traits within communities, it is
unclear how general these ﬁndings are. Several major eﬀorts
to accumulate and coordinate extensive trait and phyloge-
netic information (e.g. Phylocom, Webb et al. 2008; TRY,
Kattge et al. 2011) have expanded the potential geographic
scope of trait-based ecology, leading to a number of recent
studies of continental to global scale variation in traits,
functional and phylogenetic diversity (Reich and Oleksyn
2004, Wright et al. 2004, Ordoñez et al. 2009, Saﬁ et al.
2011, Huang et al. 2012, Swenson et al. 2012b). e
next step is to use these broad-scale patterns to evaluate the
generality of community-scale ecological theory. Doing
so requires a dataset encompassing a wide variety of species,
communities, and processes that inﬂuence these communi-
ties. It is uncertain if the inherent heterogeneity of such a
dataset will obscure any general signals, or whether hypo-
thesized ‘rules’, such as the SDH, are strong enough to be
observed regardless (Lawton 1999).
Here we test whether patterns of phylogenetic and trait
diversity in 53 439 tree communities in the eastern United
States are consistent with shifts from environmental ﬁlter-
ing to competition predicted by the SDH. In doing so, we
assess the general applicability of this hypothesis to eastern
North American forests and evaluate the utility of broad-
scale trait diversity patterns for understanding processes that
structure communities. By using a data set with a small spa-
tial grain and large spatial extent we can examine whether
community-level processes are general across continental to
Phylogenetic diversity and multivariate metrics of func-
tional diversity integrate over many diﬀerent organismal
attributes, and are potentially inﬂuenced by many diﬀerent
processes in addition to competition and environmental ﬁl-
tering, including dispersal limitation, positive interactions,
and predation or parasitism (Cavender-Bares et al. 2009,
Pavoine and Bonsall 2011, Spasojevic and Suding 2012). As
such, examining functional diversity based on single traits
that are directly related to an organism’s competitive or
stress tolerance abilities may provide less ambiguous infor-
mation about the importance of the two processes of inter-
est (Weiher et al. 1998, Swenson and Enquist 2009,
Spasojevic and Suding 2012). e expected response of a
trait depends on the trait’s ecological role (Fig. 1). In the
context of the SDH, environmental ﬁltering acts on traits
that are important for stress tolerance, favoring convergence
to an optimal trait value. is lowers within-community
trait variation, which we refer to as ‘trait diversity’.
Competition can have opposite eﬀects on traits related to
niche diﬀerences versus traits related to competitive ability
(Mayﬁeld and Levine 2010). It is expected to increase the
diversity of traits involved in resource partitioning, but
lower the diversity of traits conferring competitive domi-
nance by favoring convergence on the trait value that leads
to greatest competitive ability (Kunstler et al. 2012).
Competition can occur throughout forest development,
and competitive pressures at diﬀerent successional stages
may select for diﬀerent phenotypes (Huston and Smith
1987). Given that the majority of eastern U.S. forests are
young (Pan et al. 2011) due to logging and extensive aban-
donment of agricultural lands in the last century (Abrams
1992, Smith et al. 2009), competitive processes operating at
the early successional phases are most likely to dominate the
trait distributions of current forests.
To assess whether the variation of phylogenetic and trait
diversity in tree communities is consistent with the stress
dominance hypothesis, we developed and tested a set of
hypotheses (Table 1) for changes in community mean trait
values, trait diversity, and phylogenetic diversity along two
stress gradients (soil nutrient availability and water avail-
ability). Our hypotheses are based on competition favoring
a fast-growth, low resource-use eﬃciency strategy in benign
environments. However, this may not be realistic in older
Figure 1. Expected shifts in trait diversity of diﬀerent trait types
under the stress-dominance hypothesis. Competition and environ-
mental ﬁltering can have diﬀerent eﬀects on the within-community
dispersion of diﬀerent types of traits. Consequently, if the relative
strength of these processes varies along a stress gradients as pre-
dicted by the SDH, the diversity of diﬀerent trait types will exhibit
diﬀerent correlations with particular environmental stressors. For
example, traits related to niche diﬀerences (resource-use traits) are
expected to exhibit high diversity in highly competitive environ-
ments due to limiting similarity, whereas traits conferring greater
competitive ability will be ﬁltered and exhibit low diversity.
Traits mediating a tradeoﬀ between competitive ability and stress-
tolerance may be ﬁltered at both ends of the gradient and reach
maximum diversity in moderate environments.
Table 1. Expected shifts in phylogenetic diversity, trait values and trait diversity along a stress gradient under the stress-dominance hypothesis.
Predictions assume that environmental ﬁlters dominate community assembly in stressful environments and that competitive ﬁlters dominate
in benign environments. Solid blue lines depict expected shifts in the mean trait value across communities. Red dashed lines depict expected
shifts in trait diversity across communities.
(strong environmental ﬁltering)
*These expectations hold only
when the phenotypes
environmental tolerances and
niche relations are
phylogenetically conserved, so
that phylogenetic distance
Phylogenetic diversity will be
high due to competitive
exclusion of species with
similar phenotypes and the
Phylogenetic diversity should be low in
the most stressful environments
because only certain clades have
evolved the adaptations necessary to
tolerate these conditions.*
Sources differ on how seed size
should be affected by stress
gradients because seed mass is
typically thought to reﬂect a
tradeoff between viability and
dispersal (Kitajima 2007).
Large seeds are advantageous
under strong competition
(see references in Leishman
et al. 2000). Therefore, mean
seed mass will be high and
seed mass diversity will be
low due to competitive
Both small and large seeds can be
advantageous in stressful
environments (Leishman 2001,
Moles and Westoby 2004). Seed
mass diversity will be high.
Leaf nitrogen content
Leaf nitrogen content reﬂects a
tradeoff between stress
tolerance and competitive
dominance (Wright et al. 2004)
Competition favors faster
growth rates, leading to high
leaf nitrogen content due to
efﬁciency. Leaf nitrogen
diversity will be low due to
Stressful environments favor high
resource use efﬁciency and should
lead to low leaf nitrogen content.
Leaf nitrogen diversity will be low
due to environmental ﬁltering.
Wood density reﬂects a tradeoff
between stress tolerance and
competitive dominance (e.g.
the wood economics spectrum;
Chave et al. 2009).
Competition favors fast growth
and consequently low wood
density. Wood density
diversity will be low due to
Low water and nutrient availability in
stressful environments favors high
resource use efﬁciency and
resistance to embolism, leading to
high wood density (Hacke et al.
2001, Martínez-Cabrera et al. 2009).
Since the xylem architecture of
conifers allows them to persist in
stressful conditions despite low wood
density relative to angiosperms
(Hacke et al. 2001), this increase in
mean wood density may not be very
pronounced. Wood density diversity
will be high because high-density
angiosperms co-occur with lower
forests where competitive exclusion has selected for trees
with a shade-tolerant phenotype, or in disturbance regu-
lated forests that do not follow a traditional successional
trajectory. We considered three physiologically important
traits that represent ecological trade-oﬀs: seed mass, leaf
nitrogen content, and wood density. Previous work has
examined geographic variation in these traits (Swenson
and Weiser 2010, Siefert et al. 2012) as well as temporal
and spatial variation in phylogenetic diversity (Potter and
Woodall 2012, Hawkins et al. 2014) in eastern U.S. for-
ests, but has not evaluated whether this variation reﬂects
changes in community assembly processes along environ-
Tree community data
is study was conducted using two complementary data-
bases of vegetation plots in the eastern United States: the
U.S. Forest Service Forest Inventory and Analysis National
program (FIA; Gray et al. 2012) and the Carolina Vegetation
Survey (CVS; Peet et al. 2012) (Fig. 2). We used data from
FIA forest plots spanning all states east of the Great Plains
(approximately 96°W longitude). e CVS data set is smaller
in extent and contains plots in North Carolina, South
Carolina, Georgia, and Florida. Although both programs
for each trait in each plot by averaging the species-level trait
values for all species present in the plot, weighted by their
relative abundance (Garnier et al. 2004, Ricotta and Moretti
2011). For CVS, relative abundance was based on percent
cover and for FIA, relative abundance was based on basal area.
We constructed a phylogeny for each data set using the
Phylomatic online software ver. 2 (Webb et al. 2008),
which provides a dendrogram resolved to the genus level
based on the Angiosperm Phylogeny Group III (Stevens
2012). Branch lengths were assigned using the BladJ
function of the Phylocom software, which assigned nodal
ages down to the family-level based on (Wikström et al.
2001). Where node ages were unavailable, the software
split known distances evenly between ageless nodes and
branch tips occurring between or after known nodes. e
Phylomatic online software provided the topology for
gymnosperms, although no nodal ages were available
so branch lengths were split evenly between each node in
the gymnosperm clade. Similar phylogenies have been
useful in evaluating ecological hypotheses about the
phylogenetic relationships among species in communities
(Cavender-Bares et al. 2006, Kembel and Hubbell 2006,
Kraft and Ackerly 2010).
Trait and phylogenetic diversity
We measured trait and phylogenetic diversity using the
abundance-weighted mean pairwise distance among
species in a plot (MPD; Clarke and Warwick 1998, Webb
2000). is is equivalent to Rao’s quadratic entropy
(Botta-Dukát 2005) which has been shown to discriminate
between community assembly processes in simulated data
(Mouchet et al. 2010) and empirical data (Ricotta and
Moretti 2011). MPD is suitable for this analysis because it is
mathematically independent of richness and robust to imbal-
anced phylogenies when detecting overdispersed and clus-
tered community assembly processes (Vellend et al. 2011).
Figure 2. Geographic distribution of plots in the FIA and CVS datasets. Plots are colored by mean annual climatic water deﬁcit.
measure all tree species occurring on plots of fairly equiva-
lent size (FIA – 672 m2; CVS – 1000 m2), the two programs
diﬀer in their sampling ideology and methodology. e goal
of the FIA program is to assess the state of United States for-
est land by surveying randomly located plots. In contrast,
CVS aims to record naturally occurring plant communities
in the southeastern U.S. and chooses plot locations that
maximize homogeneity within plots and exclude potentially
human-introduced elements. We used only the most
recent survey data from each plot and excluded FIA plots
with evidence of tree planting or cutting and CVS plots
labeled as early successional plots. We conducted parallel
analyses on the two datasets separately, since diﬀerences in
sampling methodology may impact ecological inference.
Criteria used to select plots and speciﬁc methods of
plot sampling are described in Supplementary material
Appendix 1. Our ﬁnal data set consisted of 51 051 FIA plots
sampled during 1997–2010 and 2388 CVS plots sampled
during 1988–2010. Data available from the Dryad Digital
Repository: , http://dx.doi.org/10.5061/dryad.m5g 7d ..
We compiled species-level mean trait data for 269 species for
three traits from primary literature sources and publicly
available trait databases (Supplementary material
Appendix 2): seed mass (average mass of 1 seed, 0.055–
16 200 mg), wood density (oven dry mass divided by green
volume, 0.24–0.89 g cm23), and leaf nitrogen content as a
percent of dry weight (0.32–3.54%). Both the CVS and
FIA datasets contained some trees that were identiﬁed
only to the genus level (26 taxa) as well as some species for
which we were unable to obtain trait data (41 species:
20 missing all three traits, of which 15 are Crataegus species);
for these cases, we used genus-level average trait values,
calculated from the species that were present in our datasets
(Supplementary material Appendix 2). Fourteen species
retained missing values due to a lack of information at the
genus level. Seed mass spanned ﬁve orders of magnitude
and was therefore log10-transformed prior to all calculations.
We calculated the community-weighted mean trait values
calendar year. D is an eﬀective measure of overall water
stress to plants because it represents the potential additional
evaporative demand not met by available water based on
energy input and precipitation (Stephenson 1998, Lutz
et al. 2010). It has also been shown to better correlate with
tree distributions than water supply measures, such as
annual precipitation (Piedallu et al. 2013). We calculated
D for each plot by intersecting plot geographic coordinates
with 30-arc-second resolution maps of long-term average
annual PET and AET (CGIAR-CSI’s Global Aridity and
PET Database and Global High-Resolution Soil-Water
Balance database, (Trabucco and Zomer 2009, 2010)),
which were generated using WorldClim temperature and
precipitation data (Hijmans et al. 2005) under the
Hargreaves model. irty-one plots (27 FIA, 6 CVS)
were excluded as probable outliers and 118 CVS plots
were excluded from D models due to missing geographic
coordinates. A subset of FIA plots (44 394 plots) were clas-
siﬁed as ‘xeric’ or ‘mesic’ within the FIA database according
to topographic position and water availability as perceived
by the survey crew. We used these classes as a local-scale
alternate measure of water stress and used a Mann–Whitney
U-test to compare mean trait values and trait diversity
between these two groups of plots.
Soil nutrient availability was calculated for CVS
plots using principle components analysis (PCA) of 23 soil
characteristics measured at each plot (Supplementary
material Appendix 1). Correlations of individual soil vari-
ables with the ﬁrst principle component indicated that it
represents a gradient from acidic, low nutrient, stressful
conditions to benign high nutrient, basic conditions (Peet
et al. 2014). We were unable to calculate soil nutrient avail-
ability for 301 CVS plots due to missing data. We did not
calculate soil nutrient availability for FIA plots because
only a small subset had associated soil data.
We tested for monotonic relationships between mean
trait values and the two stress gradients by ﬁtting two mod-
els: a simple linear regression and a power function of the
form y axb using non-linear least-squares (chosen over lin-
ear regression on log-transformed data so that models could
be compared using AIC). We used the same models to test
for a monotonic relationship between PD and the stress gra-
dients. However, for trait diversity we had hypothesized
hump-shaped relationships so we also tested a linear model
with a quadratic term. Models of trait and phylogenetic
diversity used z-scores as the response variable. All models
were ﬁt in R ver. 2.14. e model with the lowest AIC is
reported, unless the diﬀerence in AIC was less than 2, in
which case the simpler model was used (Supplementary
material Appendix 4, Table A4.1). Because of the large
number of plots included in the analysis, all slopes diﬀered
from zero with p , 0.001, so we only report relationships
explaining at least 5% of the total variation (r2 . 0.05).
CVS phylogenetic diversity (PD) was negatively correlated
with soil nutrient availability (r 20.44; Fig. 3), the
For trait diversity, traits were standardized by their mean
and standard deviations across species and then distances
among species were computed as the Euclidean distance
between these values. Species with missing trait values
were omitted from calculations involving the missing trait
(aﬀecting 728 plots, but with only 2.3% of the total cover
in these plots omitted). For phylogenetic diversity, the dis-
tance between species is the total branch length between
them on the phylogeny. Calculations were performed in R
ver. 2.14.0 (R Core Team) using the FD (Laliberté and
Shipley 2011) and picante (Kembel et al. 2010) packages.
Inference of environmental ﬁltering and competition is
usually based on the deviation (z-score) of a community’s
functional diversity from the diversity value expected under
a null model that simulates random assembly (Cornwell
et al. 2006, Mouillot et al. 2007, Swenson and Enquist
2007). In addition, z-scores allow diversity to be compared
among communities that diﬀer in species number, since
MPD can be correlated with species richness due to sam-
pling eﬀects (Weiher 2011). We generated trait diversity
null distributions for each plot by randomly shuﬄing
trait values across the entire species pool in each data set
1000 times and recalculating trait diversity each time
(Swenson and Weiser 2010). We then calculated trait diver-
sity z-scores by subtracting the mean of the null distribution
from the observed trait diversity and dividing by the stan-
dard deviation of the null distribution. Plots falling in the
95th or higher percentile of the null distribution were
considered ‘overdispersed’, exhibiting higher diversity than
expected by random assembly, and plots in the 5th or lower
percentile were considered ‘underdispersed’, exhibiting
lower diversity than expected under random assembly
(Swenson et al. 2012a). A similar model was used to
generate null distributions of phylogenetic diversity for each
plot, except in this case, species were shuﬄed across the
tips of the phylogeny. Because our phylogeny had an
unbalanced, decelerating topology resulting from the initial
gymnosperm-angiosperm bifurcation, we also calculated
MPD using only angiosperm taxa in order to examine
potential inconsistencies. We also evaluated the mean near-
est taxon distance (MNTD) which may be less sensitive to
the angiosperm-gymnosperm split (Supplementary material
Appendix 3). Because unconstrained null models can be
biased toward identifying underdispersion (de Bello et al.
2012), we also calculated z-scores using null models in
which the species at each site were randomly drawn from
the set of species with environmental niches spanning the
environmental conditions found at that site. Results based
on this constrained null model were qualitatively similar
and are addressed in the Discussion section.
Environmental data and models
We chose to examine two of the most important environ-
mental variables known to structure plant communities
worldwide: soil nutrient availability and water availability
(Archibold 1995). To represent water stress, we used aver-
age annual climatic water deﬁcit (D) (Stephenson 1990),
deﬁned as the diﬀerence between potential evapotrans-
piration (PET) and actual evapotranspiration (AET) over a
Figure 3. Phylogenetic diversity in CVS and FIA plots along water deﬁcit and soil nutrient availability gradients. Phylogenetic diversity is
measured as the mean pair-wise phylogenetic distance between taxa in a community. Positive z-score values indicate higher diversity
and negative values indicate lower diversity relative to a null model of random community assembly with respect to phylogenetic relation-
ships. Opaque points are above the 95th or below the 5th percentile of the null distribution and points are colored by the proportion of
the community that is comprised of angiosperm taxa. Regression lines are shown for relationships with r2
. 0.05. Horizontal bands of
color indicate that phylogenetic diversity of a community is largely driven by the relative abundance of gymnosperms versus angiosperms
in a community which results from the deep initial split between these clades.
opposite of our prediction that fertile sites should exhibit
phylogenetic overdispersion due to stronger competition
and weak environmental ﬁltering. PD was not correlated
with water deﬁcit in either data set (Table 2). For both CVS
and FIA, PD was strongly inﬂuenced by the presence of
gymnosperms, increasing as the proportion of gymnosperms
in the community increased (Fig. 3). PD changed dramati-
cally when only angiosperm taxa were included in the analy-
sis, eliminating the previously observed negative correlation
between soil nutrient availability and PD (Supplementary
material Appendix 3, Fig. A3.2). Other diversity metrics
performed similarly (Supplementary material Appendix 3).
Community-weighted mean trait values
Mean leaf nitrogen content and wood density responded as
predicted to the stress gradients. However mean seed
mass increased with environmental stress, the opposite of
our initial hypothesis (Fig. 4 and 5, Table 2). Most relation-
ships were weak, explaining less than 10% of the total varia-
tion in mean trait values. e strongest relationship was in
CVS plots between mean leaf nitrogen and soil nutrient
0.38), where leaf nitrogen content initially
increased with soil nutrient availability and reached a plateau
at high levels (Fig. 5). Other model results are in Table 2.
Analysis of mean traits using the local-scale xeric-mesic cat-
egorization yielded trends consistent with the water deﬁcit
models. Xeric sites had signiﬁcantly higher wood density
and seed mass and lower leaf nitrogen content than mesic
sites (Supplementary material Appendix 5, Fig. A5.1).
Trait diversity showed no clear relationship (r² , 0.05) with
either stress gradient, with two exceptions (Fig. 4, 5). We
found a moderately weak, negative relationship between seed
mass diversity and water deﬁcit in CVS plots (r² 0.09). We
also detected a weak quadratic relationship between wood
density diversity and soil nutrient availability (Fig. 5) with
diversity reaching a minimum in the middle of the gradient.
Trait diversity was higher in xeric than in mesic FIA plots for
seed mass and wood density (p , 0.001), however, these dif-
ferences were small (Supplementary material Appendix 5,
Fig. A5.1) and may not be biologically meaningful.
We found very little overdispersion of trait diversity in
CVS and FIA plots (Table 3) and this may have decreased
our ability to detect the hypothesized shifts in trait diversity
along stress gradients. Although in several cases peak
diversity appears to occur in the middle of the stress gradient
(Fig. 4, 5), permutation tests revealed that the distribution
Table 2. Models relating mean traits, trait diversity and phylogenetic diversity to water and soil nutrient availability. Models in bold highlight
AIC supported models explaining at least 5% of the variation. r2 for non-linear power models were calculated using the residual sum of
squares (deviance) according to 1 2 (SSresidual/SStotal) (Kvålseth 1985). The estimate reported is the slope parameter for linear models, the
quadratic parameter for quadratic models, and the exponential parameter for power models. N is the number of plots used in each model.
Predictor Response Dataset Form r2Estimate Std. Err. tp N
Water deﬁcit Seed mass FIA linear 0.10 3.54E-03 4.68E-05 75.6 0.00E 00 51023
CVS linear 0.03 7.27E-04 9.24E-05 7.9 5.76E-15 2264
Wood density FIA linear 0.07 2.38E-04 3.77E-06 63.2 0.00E 00 51023
CVS linear 0.08 1.53E-04 1.06E-05 14.5 2.31E-45 2264
Nitrogen % FIA linear 0.01 23.49E-04 1.61E-05 221.6 1.95E-103 51023
CVS linear 0.09 29.99E-04 6.63E-05 215.1 5.96E-49 2264
Seed mass diversity FIA quadratic 0.01 28.95E-06 3.61E-07 224.8 2.79E-135 50501
CVS quadratic 0.09 3.35E-06 1.33E-06 2.5 1.18E-02 2256
Wood density diversity FIA quadratic 0.02 21.43E-06 2.66E-07 25.4 7.99E-08 50501
CVS quadratic 0.03 4.61E-06 9.17E-07 5.0 5.39E-07 2256
Nitrogen % diversity FIA quadratic 0.02 2.00E-06 2.66E-07 7.5 5.34E-14 50501
CVS linear 0.02 9.65E-04 1.27E-04 7.6 5.53E-14 2257
Phylogenetic diversity FIA power 0.00 5.78E-01 2.61E-01 2.2 2.67E-02 50501
CVS power 0.02 3.41E-01 8.58E-02 4.0 7.15E-05 2257
Seed mass CVS power 0.00 23.16E-02 1.21E-02 22.6 9.18E-03 2087
Wood density CVS power 0.07 26.88E-02 5.32E-03 212.9 7.77E-37 2087
Nitrogen % CVS power 0.38 3.53E-01 1.04E-02 34.1 1.40E-202 2087
Seed mass diversity CVS quadratic 0.03 21.49E-01 1.92E-02 27.7 1.61E-14 2079
Wood density diversity CVS quadratic 0.05 8.77E-02 1.29E-02 6.8 1.20E-11 2079
Nitrogen % diversity CVS quadratic 0.04 9.11E-02 1.36E-02 6.7 2.51E-11 2080
Phylogenetic diversity CVS linear 0.19 25.57E-01 2.51E-02 222.2 3.34E-98 2080
of overdispersed plots along the stress gradient did not
diﬀer from the distribution of non-overdispersed plots.
Overdispersed plots appear to occur in the middle of the
gradient simply because most plots occur in the middle of
Our analysis of phylogenetic diversity clearly demonstrates
the importance of taxonomic scale for interpreting phyloge-
netic overdispersion. Analyzing communities containing
both angiosperms and gymnosperms necessitates a deep ini-
tial bifurcation in any phylogeny which leads to phylo-
genetic diversity being chieﬂy driven by the ratio of
angiosperm and gymnosperm taxa. Because these two groups
are not as functionally and ecologically distinct as this bifur-
cation would imply, phylogenetic diversity is a poor proxy
for functional diversity. is dependence of phylogenetic
diversity on taxonomic breadth of the phylogeny is well-
known (Cavender-Bares et al. 2006, Vellend et al. 2011),
and our work suggests that measures of phylogenetic diver-
sity are diﬃcult to interpret in a functional context when a
community includes both angiosperms and gymnosperms.
However, we chose not to interpret phylogenetic diversity
from only the angiosperm portion of the community because
doing so eliminates a functionally non-random subset and
could mis-represent the process of community assembly.
Mean trait values
Weak shifts in community-weighted mean trait values
along the two stress gradients provide some evidence that
ﬁlters act to shape tree communities along these gradients.
Sites with high water deﬁcit, where potential evaporative
demand is much higher than water availability, tended to
have species with lower nitrogen content in their leaves,
denser wood, and larger seeds. From the SDH, we predicted
all but the last relationship, hypothesizing that higher stress
environments with lower resource availability favor plants
with higher resource-use eﬃciency, whereas if low stress
environments are structured by competition, plants with
lower resource-use eﬃciency, but faster growth will be
e strongest relationship we observed was between soil
nutrient availability and leaf nitrogen content, which is con-
sistent with previous studies (Ordoñez et al. 2009). Because
we observe a response in literature-based species-level
mean traits, our analysis provides evidence of an environ-
mental ﬁlter rather than a plastic response of individuals to
local conditions or soil enrichment by decomposition of
high nitrogen-content leaf litter. e decrease in leaf
nitrogen content with increasing water stress that we
observed in CVS plots was consistent with our hypothesis
that greater resource use eﬃciency would be promoted in
stressful environments. is is contrary to studies showing
increased leaf nitrogen in arid environments, as an adapta-
tion to prevent water loss by allowing stomata to remain
closed for longer periods of time (Wright et al. 2001, 2005).
However, these studies included sites which were more
arid than the climate of eastern North America.
Among the three traits we examined, seed mass showed
the least response to both stress gradients. is may reﬂect
the fact that seed mass is tied to dispersal strategy (Leishman
2001, Kitajima 2007), which we do not expect to be strongly
inﬂuenced by either of the two stress gradients. e notable
positive relationship between seed mass and water deﬁcit
runs counter to our initial prediction that large seeds would
Figure 4. Eﬀect of water stress on community-weighted mean trait values and trait diversity in FIA and CVS plots. Water deﬁcit is plotted on the x-axis with higher water deﬁcit corresponding
to higher water stress. Panel (A) shows mean trait values while panel (B) show trait diversity z-scores. Positive z-score values indicate high diversity and negative values indicate low diversity compared
to a null model of random community assembly with respect to traits. Black points are above the 95th or below the 5th percentile of the null distribution, whereas grey points are between these
percentiles. Lines show the best ﬁt models and only included if r2 . 0.05 (Table 2 and Supplementary material Appendix 4, Table A4.1).
Figure 5. Eﬀect of soil nutrient availability on community-weighted mean trait values and trait diversity in CVS plots. e x-axis is the ﬁrst
principle component of a PCA of 23 soil variables and represents a soil nutrient availability gradient ranging from acidic, stressful
conditions (negative values) to basic, benign conditions (positive values). e ﬁrst column shows mean trait values and the second
column shows trait diversity z-scores, as described in Fig. 4. Black points are above the 95th or below the 5th percentile of the null distribu-
tion, whereas grey points are between these percentiles. Lines show the best-ﬁt models and are only included if r2 . 0.05 (Table 2 and
Supplementary material Appendix 4, Table A4.1).
be competitively superior in low stress environments.
Instead it seems to support experimental evidence that large
seeds are advantageous in drier soil because they confer
greater seedling survival (see references in Leishman et al.
2000). Given the slight trend toward lower seed mass diver-
sity at higher water deﬁcit, our data suggest that the rela-
tionship between seed mass and water availability may be
driven by ﬁltering for larger seeds at drier sites. We ﬁnd
no evidence for competitive ﬁlters on seed mass in benign
e observed shifts in mean trait values diﬀer from
those reported previously in Forest Inventory and Analysis
plots, in which annual precipitation was positively corre-
lated with seed mass and wood density and negatively
correlated with leaf nitrogen content (Swenson and
Weiser 2010). is apparent disagreement can be resolved
by recognizing that annual precipitation measures water
supply whereas water deﬁcit measures evaporative demand.
In fact, annual precipitation and water deﬁcit were weakly
positively correlated along our stress gradient (r 0.20),
measured (Leishman et al. 2000, Sungpalee et al. 2009,
Albert et al. 2010, Auger and Shipley 2013). However, a
study examining community-scale processes, regardless of
spatial extent, may still need to account for local variation in
traits (Albert et al. 2011).
e trait-diversity z-scores that we analyzed are known
to be susceptible to the formulation of the null model
(Mouchet et al. 2010, de Bello 2012). Unconstrained
null models like ours are biased toward detecting underdis-
persion because regional species pools may diﬀer in their
trait distributions. Our null model implicitly assumed that
any species could colonize any site. If certain geographic
areas do not contain species with trait values covering the
entire range of trait values found in the total species pool,
then our null model would bias sites in those areas toward
underdispersion. However, we checked the range of trait
values that occurred within equal-area grid cells across our
study region and found no geographic bias in these trait
ranges. Unconstrained null models are also biased toward
underdispersion because environmental ﬁlters operate prior
to biotic interactions so that observed communities will
typically have lower trait diversity than expected of a com-
munity randomly assembled from species across diﬀerent
environments or habitats (de Bello et al. 2012). One solu-
tion is to attempt to remove abiotic environmental ﬁlters by
comparing communities to a null expectation acquired
only from species that could potentially tolerate a site’s envi-
ronment (Peres-Neto et al. 2001, de Bello et al. 2012).
While this approach does not allow us to compare the rela-
tive inﬂuence of environmental ﬁltering and competition,
which is crucial for testing the SDH, we wanted to aﬃrm
that the underdispersion and lack of systematic variation
that we observed was not an artifact of an unconstrained
null model. We re-calculated trait diversity z-scores for
CVS plots and a subset of southeastern FIA plots using
environmentally constrained null models that only permit-
ted shuﬄing of species among sites that fell within species’
environmental niches. Although this resulted in a small
increase in the number of plots exhibiting trait diversity
overdispersion, there was no change in the lack of observed
relationships between trait diversity and environmental
gradients (Supplementary material Appendix 6). Analytical
approaches that separate the eﬀects of competition from
environmental ﬁltering (de Bello et al. 2012) are especially
useful when these processes are predicted to ﬁlter traits
toward similar values. In our case, competitive and environ-
mental ﬁlters were expected to select for diﬀerent trait
e overall lack of plots exhibiting trait overdispersion
limited our ability to discern shifts in trait diversity
along the gradients. Yet, the pervasiveness of trait diversity
underdispersion may be ecologically meaningful. Several
other studies have found consistent underdispersion in
plant communities along environmental gradients. De Bello
et al. (2009) attributed underdispersion in speciﬁc leaf
area throughout a moisture gradient to environmental ﬁl-
tering. Savage and Cavender-Bares (2012) also found that
environmental ﬁltering was important for willow tree
communities along the length of a hydrologic gradient,
with trees at the dry end exhibiting traits associated with
drought tolerance and trees at the wet end exhibiting traits
with highest water deﬁcit occurring in locations with
moderate annual precipitation.
Trait diversity did not notably respond to either of the stress
gradients we examined. is can be interpreted in several
ways: 1) our data set did not encompass a wide enough envi-
ronmental range to capture both stressful and benign condi-
tions, 2) the traits we examined are not inﬂuenced by
the environmental gradients measured, 3) the species-level
mean trait values we used masked local trait–environment
relationships, 4) our metrics did not accurately capture exist-
ing trait convergence or divergence, or 5) our hypotheses
about processes structuring tree communities along stress
gradients are not generally true across eastern North
It is unlikely our dataset failed to encompass a viable
stress gradient or that the traits we examined were not inﬂu-
enced by it, because we do observe shifts in mean trait
values along both environmental gradients, as have others
(Wright et al. 2005, Swenson and Weiser 2010). Combined
with the signiﬁcant trait underdispersion that we observe in
many plots, this suggests that both gradients encompass con-
ditions stressful enough to impose ﬁlters (albeit weak) on
It is possible that our metrics did not accurately
capture existing patterns of trait diversity, either by ignoring
intraspeciﬁc variation or by our choice of diversity metric
and null model. Using species-level mean traits may have
masked local mechanisms whereby trait plasticity among
individuals allows coexistence through niche partitioning
(Clark 2010, Burns and Strauss 2012). Several studies have
found trait divergence in local communities when account-
ing for intraspeciﬁc trait variation (Jung et al. 2010, de Bello
et al. 2011, Paine et al. 2011). e necessity of including
intraspeciﬁc trait variation in large-scale studies has been
debated, since for many traits, variation between species is
usually greater than variation within species when enough
species are included. is is likely true for the three traits we
Table 3. Proportion of FIA and CVS plots with signiﬁcantly overdis-
persed or underdispersed trait diversity. Overdispersed plots have
trait diversity above the 95th percentile of the null distribution,
underdispersed plots are below the 5th percentile, and random
plots are between the 5th and 95th percentiles. More plots are
under dispersed than overdispersed, but in general most plots have
a level of trait diversity that does not differ from random assembly.
FIA wood density
1.33 75.42 23.25
CVS wood density
0.93 70.41 28.67
FIA leaf nitrogen
1.35 82.69 15.96
CVS leaf nitrogen
2.03 85.82 12.15
FIA seed mass
3.18 81.71 15.10
CVS seed mass
7.95 88.43 3.62
communities, although it may operate in more restricted
e broad geographic extent and large number of com-
munities in our analysis spanned a variety of climates, habi-
tat types, successional stages, and disturbance regimes.
is heterogeneity of environmental contexts is a necessary
condition for testing the generality of a theory in commu-
nity ecology, but it also could have obscured patterns result-
ing from the SDH if this hypothesis only applies under
certain circumstances. e majority of the plots that we
analyzed were embedded in a human-modiﬁed landscape
and occurred at a range of successional stages. is may
have masked trait–environment relationships, given that
the importance of dispersal limitation, abiotic ﬁlters,
and biotic interactions are known to shift throughout forest
succession as are the traits that are aﬀected by these pro-
cesses (Prach et al. 1997, Douma et al. 2012, Kröber
et al. 2012 and references therein). Additionally, the SDH
may not apply across forests experiencing diﬀerent levels of
disturbance, since disturbance-related ﬁlters on tree traits
can vary across disturbance regimes (Loehle 2000). Future
studies could assess whether SDH-related trait variation is
more evident when restricting analyses to particular eco-
logical contexts. Given the contingent nature of many eco-
logical systems, this approach could aid the search for
general principles in a time of increasing data availability
Acknowledgements – is work resulted from a Dimensions of
Biodiversity Distributed Graduate Seminar at the Univ. of North
Carolina at Chapel Hill and we thank the participants for compil-
ing the trait data and initiating the idea for project: K. Becraft,
C. Fieseler, C. Hakkenberg, C. Mitchell, C. Payne, K. Peck,
D. Tarasi, and C. Urbanowicz. We also thank the entire DBDGS
community for their support. is project would not be possible
without the individuals who collected the FIA and CVS data
and made it available. Special thanks to Robert Peet and Nathan
Swenson for the use of their data and to four reviewers and the
subject editor whose comments signiﬁcantly improved the manu-
script. is work was funded by NSF grant #DEB-1050680 to the
Univ. of Washington (J. Parrish and S. Andelman, PIs) through a
subcontract to the Univ. of North Carolina at Chapel Hill (AHH,
C. Mitchell and R. Peet, PIs). All authors contributed equally to
Abrams, M. 1992. Fire and the development of oak forests.
– Bioscience 42: 346–353.
Albert, C. H. et al. 2010. A multi-trait approach reveals the struc-
ture and the relative importance of intra- vs. interspeciﬁc
variability in plant traits. – Funct. Ecol. 24: 1192–1201.
Albert, C. H. et al. 2011. When and how should intraspeciﬁc
variability be considered in trait-based plant ecology?
– Perspect. Plant Ecol. Evol. Syst. 13: 217–225.
Archibold, O. W. 1995. Ecology of world vegetation. – Chapman
Auger, S. and Shipley, B. 2013. Inter-speciﬁc and intra-speciﬁc trait
variation along short environmental gradients in an old-
growth temperate forest. – J. Veg. Sci. 24: 419–428.
Botta-Dukát, Z. 2005. Rao’s quadratic entropy as a measure of
functional diversity based on multiple traits. – J. Veg. Sci.
associated with ﬂooding tolerance. e persistent underdis-
persion in our forest plots may indicate a role for environ-
mental ﬁltering across many habitats. Our results do imply
that, at this large spatial extent, competition is unlikely to
strongly inﬂuence tree community trait structure, since
we found no evidence of competitive ﬁltering along either
stress gradient. e relationships between community mean
leaf nitrogen and wood density could have been driven
principally by environmental ﬁltering at the stressful end
of the gradient even without competition occurring in
benign environments. Additionally, we did not observe the
predicted shift in mean seed mass, which relied on com-
petitive ﬁlters acting at the benign end of the gradient.
Instead, we observed a pattern consistent with environmen-
tal ﬁltering for larger seeds in dry environments.
Overall, trait diversity showed some support for the ﬁrst
part of the SDH, environmental ﬁltering along stress
gradients, but no support for predictions based on the
importance of competition. is may be because plant traits
are poor indicators of competition among trees. Traits in
local tree communities result from competition integrated
over the entire lifespan of the trees and reﬂect selective mor-
tality at several stages of forest development. Competition
for light at the establishment phase may favor traits associ-
ated with rapid growth, whereas long-term population-
scale competition in closed-canopy forests selects for
shade-tolerance and higher resource-use eﬃciency (Huston
and Smith 1987). Using trait data to make inferences about
competition in forests may prove most feasible when analy-
ses are restricted to a narrowly deﬁned successional stage. An
alternative interpretation of our results is that many tree
communities are assembled stochastically with respect to
the traits we measured, which would be supported by the
lack of trait under or overdispersion observed in the majority
of plots. It is possible that examining other traits would
yield diﬀerent results and support the existence of competi-
tive ﬁlters in less stressful environments.
Generality of the stress-dominance hypothesis
e stress-dominance hypothesis appears logically simple:
communities in stressful environments should be struc-
tured by environmental ﬁlters, whereas competition should
be of relatively greater importance in less stressful environ-
ments. However, we detected no strong evidence to sup-
port this hypothesis in either of the two datasets examined.
One beneﬁt of utilizing two distinct datasets was that we
were able to qualitatively evaluate whether diﬀerences in
sampling ideology and methodology inﬂuenced our ability
to detect processes structuring communities. We expected
that the greater within-site heterogeneity present in the
FIA plots relative to the CVS plots would result in stron-
ger trait–environment relationships in the CVS dataset.
However, both datasets were equally able to detect rela-
tionships between traits and water deﬁcit and, when pres-
ent, these relationships were congruent. From either
dataset, we would conclude that the best interpretation
of the weak shifts in trait values and unresponsiveness of
trait diversity to environmental gradient is that the SDH
is not a general paradigm for eastern North American tree
HilleRisLambers, J. et al. 2012. Rethinking community assembly
through the lens of coexistence theory. – Annu. Rev. Ecol.
Evol. Syst. 43: 227–248.
Huang, S. et al. 2012. Traits, trees and taxa: global dimensions of
biodiversity in mammals. – Proc. R. Soc. B 279: 4997–5003.
Huston, M. and Smith, T. 1987. Plant succession: life history and
competition. – Am. Nat. 130: 168–198.
Jung, V. et al. 2010. Intraspeciﬁc variability and trait-based
community assembly. – J. Ecol. 98: 1134–1140.
Kattge, J. et al. 2011. TRY – a global database of plant traits.
– Global Ecol. Biogeogr. 17: 2905–2935.
Keddy, P. A. 1992. Assembly and response rules: two goals for
predictive community ecology. – J. Veg. Sci. 3: 157–164.
Kembel, S. W. and Hubbell, S. P. 2006. e phylogenetic structure
of a neotropical forest tree community. – Ecology 87:
Kembel, S. W. et al. 2010. Picante: R tools for integrating
phylogenies and ecology. – Bioinformatics 26: 1463–1464.
Kitajima, K. 2007. Seed and seedling ecology. – In: Pugnaire, F. I.
and Valladares, F. (eds), Functional plant ecology, 2nd ed.
CRC Press, pp. 549–579.
Kluge, J. and Kessler, M. 2011. Phylogenetic diversity, trait
diversity and niches: species assembly of ferns along a tropical
elevational gradient. – J. Biogeogr. 38: 394–405.
Kraft, N. and Ackerly, D. 2010. Functional trait and phylogenetic
tests of community assembly across spatial scales in an
Amazonian forest. – Ecol. Monogr. 80: 401–422.
Kraft, N. et al. 2008. Functional traits and niche-based tree
community assembly in an Amazonian forest. – Science
Kröber, W. et al. 2012. Leaf trait–environment relationships
in a subtropical broadleaved forest in south-east China.
– PLoS One 7: e35742.
Kunstler, G. et al. 2012. Competitive interactions between forest
trees are driven by species’ trait hierarchy, not phylogenetic
or functional similarity: implications for forest community
assembly. – Ecol. Lett. 15: 831–840.
Kvålseth, T. 1985. Cautionary note about R2. – Am. Stat.
Laliberté, E. and Shipley, B. 2011. FD: measuring functional
diversity from multiple traits, and other tools for functional
ecology. – R package ver. 1.0-11.
Lawton, J. H. 1999. Are there general laws in ecology? – Oikos
Leishman, M. R. 2001. Does the seed size/number trade-oﬀ model
determine plant community structure? An assessment of the
model mechanisms and their generality. – Oikos 93: 294–302.
Leishman, M. R. et al. 2000. e evolutionary ecology of seed size.
– In: Fenner, M. (ed.), Seeds: the ecology of regeneration in
plant communities, 2nd ed. CAB International, pp. 31–58.
Loehle, C. 2000. Strategy space and the disturbance spectrum:
a life-history model for tree species coexistence. – Am. Nat.
Lutz, J. A. et al. 2010. Climatic water deﬁcit, tree species
ranges, and climate change in Yosemite National Park.
– J. Biogeogr. 37: 936–950.
Machac, A. et al. 2011. Elevational gradients in phylogenetic struc-
ture of ant communities reveal the interplay of biotic and
abiotic constraints on diversity. – Ecography 34: 364–371.
Martínez-Cabrera, H. I. et al. 2009. Wood anatomy and
wood density in shrubs: responses to varying aridity along
transcontinental transects. – Am. J. Bot. 96: 1388–1398.
Mason, N. W. H. et al. 2012. Changes in coexistence mechanisms
along a long-term soil chronosequence revealed by functional
trait diversity. – J. Ecol. 100: 678–689.
Mayﬁeld, M. M. and Levine, J. M. 2010. Opposing eﬀects of
competitive exclusion on the phylogenetic structure of
communities. – Ecol. Lett. 13: 1085–1093.
Bryant, J. A. et al. 2008. Microbes on mountainsides: contrasting
elevational patterns of bacterial and plant diversity. – Proc.
Natl Acad. Sci. USA 105: 11505–11511.
Burns, J. and Strauss, S. 2012. Eﬀects of competition on
phylogenetic signal and phenotypic plasticity in plant func-
tional traits. – Ecology 93: 126–137.
Cadotte, M. W. et al. 2008. Evolutionary history and the eﬀect of
biodiversity on plant productivity. – Proc. Natl Acad. Sci. USA
Cavender-Bares, J. et al. 2006. Phylogenetic structure of
Floridian plant communities depends on taxonomic and
spatial scale. – Ecology 87: S109–S122.
Cavender-Bares, J. et al. 2009. e merging of community ecology
and phylogenetic biology. – Ecol. Lett. 12: 693–715.
Chave, J. et al. 2009. Towards a worldwide wood economics
spectrum. – Ecol. Lett. 12: 351–366.
Clark, J. 2010. Individuals and the variation needed for high
species diversity in forest trees. – Science 327: 1129–1132.
Clarke, K. R. and Warwick, R. M. 1998. A taxonomic distinctness
index and its statistical properties. – J. Appl. Ecol. 35:
Cornwell, W. K. et al. 2006. A trait-based test for habitat ﬁltering:
convex hull volume. – Ecology 87: 1465–1471.
de Bello, F. 2012. e quest for trait convergence and divergence
in community assembly: are null-models the magic wand?
– Global Ecol. Biogeogr. 21: 312–317.
de Bello, F. et al. 2009. Partitioning of functional diversity reveals
the scale and extent of trait convergence and divergence.
– J. Veg. Sci. 20: 475–486.
de Bello, F. et al. 2011. Quantifying the relevance of intraspeciﬁc
trait variability for functional diversity. – Methods Ecol. Evol.
de Bello, F. et al. 2012. Functional species pool framework to
test for biotic eﬀects on community assembly. – Ecology 93:
Douma, J. C. et al. 2012. Succession-induced trait shifts across
a wide range of NW European ecosystems are driven by
light and modulated by initial abiotic conditions. – J. Ecol.
Emerson, B. C. and Gillespie, R. G. 2008. Phylogenetic analysis
of community assembly and structure over space and time.
– Trends Ecol. Evol. 23: 619–630.
Garnier, E. et al. 2004. Plant functional markers capture ecosystem
properties during secondary succession. – Ecology 85:
Gotelli, N. J. and Graves, G. R. 1996. Null models in ecology.
– Smithsonian Inst. Press.
Graham, C. H. and Fine, P. V. A. 2008. Phylogenetic beta diversity:
linking ecological and evolutionary processes across space in
time. – Ecol. Lett. 11: 1265–1277.
Graham, C. et al. 2012. Untangling the inﬂuence of ecological and
evolutionary factors on trait variation across hummingbird
assemblages. – Ecology 93: S99–S111.
Gray, A. N. et al. 2012. Forest Inventory and Analysis Database
of the United States of America (FIA). – Biodivers. Ecol. 4:
Grime, J. 1977. Evidence for the existence of three primary strate-
gies in plants and its relevance to ecological and evolutionary
theory. – Am. Nat. 111: 1169–1194.
Hacke, U. G. et al. 2001. Trends in wood density and structure
are linked to prevention of xylem implosion by negative
pressure. – Oecologia 126: 457–461.
Hawkins, B. A. et al. 2014. Community phylogenetics at the
biogeographical scale: cold tolerance, niche conservatism and
the structure of North American forests – J. Biogeogr. 41:
Hijmans, R. J. et al. 2005. Very high resolution interpolated climate
surfaces for global land areas. – Int. J. Clim. 25: 1965–1978.
Sungpalee, W. et al. 2009. Intra- and interspeciﬁc variation in wood
density and ﬁne-scale spatial distribution of stand-level
wood density in a northern ai tropical montane forest.
– J. Trop. Ecol. 25: 359–370.
Swenson, N. and Enquist, B. 2007. Ecological and evolutionary
determinants of a key plant functional trait: wood density and
its community-wide variation across latitude and elevation.
– Am. J. Bot. 94: 451–459.
Swenson, N. G. and Enquist, B. J. 2009. Opposing assembly
mechanisms in a Neotropical dry forest: implications for
phylogenetic and functional community ecology. – Ecology
Swenson, N. G. and Weiser, M. D. 2010. Plant geography
upon the basis of functional traits: an example from eastern
North American trees. – Ecology 91: 2234–2241.
Swenson, N. G. et al. 2012a. Temporal turnover in the com-
position of tropical tree communities: functional determinism
and phylogenetic stochasticity. – Ecology 93: 490–499.
Swenson, N. G. et al. 2012b. e biogeography and ﬁltering of
woody plant functional diversity in North and South
America. – Global Ecol. Biogeogr. 21: 798–808.
Tilman, D. 2004. Niche tradeoﬀs, neutrality, and community
structure: a stochastic theory of resource competition, invasion,
and community assembly. – Proc. Natl Acad. Sci. USA 101:
Trabucco, A. and Zomer, R. J. 2009. Global Aridity Index (Global-
Aridity) and Global Potential Evapo-Transpiration (Global-
PET) Geospatial Database. – CGIAR Consort. Spat. Inf.
Trabucco, A. and Zomer, R. J. 2010. Global Soil Water Balance
Geospatial Database. – CGIAR Consort. Spat. Inf.
Vellend, M. et al. 2011. Measuring phyologenetic biodiversity.
– In: Magurran, A. E. and McGill, B. J. (eds), Biological
diversity: frontiers in measurement and assessment. Oxford
Univ. Press, pp. 194–207.
Webb, C. 2000. Exploring the phylogenetic structure of ecological
communities: an example for rain forest trees. – Am. Nat. 156:
Webb, C. O. et al. 2002. Phylogenies and community ecology.
– Annu. Rev. Ecol. Syst. 33: 475–505.
Webb, C. O. et al. 2008. Phylocom: software for the analysis
of phylogenetic community structure and trait evolution.
– Bioinformatics 24: 2098–2100.
Weiher, E. 2011. A primer of trait and functional diversity. – In:
Magurran, A. E. and McGill, B. J. (eds), Biological diversity:
frontiers in measurement and assessment. Oxford Univ. Press,
Weiher, E. and Keddy, P. 1995. Assembly rules, null models, and
trait dispersion: new questions from old patterns. – Oikos
Weiher, E. et al. 1998. Community assembly rules, morphological
dispersion, and the coexistence of plant species. – Oikos
Wikström, N. et al. 2001. Evolution of the angiosperms:
calibrating the family tree. – Proc. R. Soc. B 268: 2211–2220.
Wright, I. J. et al. 2001. Strategy shifts in leaf physiology, structure
and nutrient content between species of high- and low-
rainfall and high- and low-nutrient habitats. – Funct. Ecol. 15:
Wright, I. J. et al. 2004. e worldwide leaf economics spectrum.
– Nature 428: 821–827.
Wright, I. J. et al. 2005. Modulation of leaf economic traits
and trait relationships by climate. – Global Ecol. Biogeogr.
Moles, A. T. and Westoby, M. 2004. Seedling survival and
seed size: a synthesis of the literature. – J. Ecol. 92: 372–383.
Mouchet, M. A. et al. 2010. Functional diversity measures: an
overview of their redundancy and their ability to discriminate
community assembly rules. – Funct. Ecol. 24: 867–876.
Mouillot, D. et al. 2007. Limiting similarity, niche ﬁltering and
functional diversity in coastal lagoon ﬁsh communities.
– Estuar. Coast Shelf Sci. 71: 443–456.
Ordoñez, J. C. et al. 2009. A global study of relationships between
leaf traits, climate and soil measures of nutrient fertility.
– Global Ecol. Biogeogr. 18: 137–149.
Paine, C. E. T. et al. 2011. Functional traits of individual trees
reveal ecological constraints on community assembly in
tropical rain forests. – Oikos 120: 720–727.
Pan, Y. et al. 2011. Age structure and disturbance legacy of
North American forests. – Biogeosciences 8: 715–732.
Pavoine, S. and Bonsall, M. B. 2011. Measuring biodiversity
to explain community assembly: a uniﬁed approach. – Biol.
Rev. 86: 792–812.
Peet, R. K. et al. 2012. Vegetation-plot database of the Carolina
Vegetation Survey. – Biodivers. Ecol. 4: 243–253.
Peet, R. K. et al. 2014. Herbaceous layer species richness of
southeastern forests and woodlands. – In: Gilliam, F. S. and
Roberst, M. R. (eds), e herbaceous layer in forests of eastern
North America. Oxford Univ. Press, pp. 255–276.
Peres-Neto, P. et al. 2001. Environmentally constrained null models:
site suitability as occupancy criterion. – Oikos 93: 110–120.
Piedallu, C. et al. 2013. Soil water balance performs better than
climatic water variables in tree species distribution modelling.
– Global Ecol. Biogeogr. 22: 470–482.
Potter, K. M. and Woodall, C. W. 2012. Trends over time in tree
and seedling phylogenetic diversity indicate regional diﬀerences
in forest biodiversity change. – Ecol. Appl. 22: 517–531.
Prach, K. et al. 1997. Changes in species traits during succession:
a search for pattern. – Oikos 79: 201–205.
Reich, P. B. and Oleksyn, J. 2004. Global patterns of plant leaf
N and P in relation to temperature and latitude. – Proc.
Natl Acad. Sci. USA 101: 11001–11006.
Ricotta, C. and Moretti, M. 2011. CWM and Rao’s
quadratic diversity: a uniﬁed framework for functional ecology.
– Oecologia 167: 181–188.
Saﬁ, K. et al. 2011. Understanding global patterns of mammalian
functional and phylogenetic diversity. – Phil. Trans. R. Soc.
B 366: 2536–2544.
Savage, J. and Cavender-Bares, J. 2012. Habitat specialization and
the role of trait lability in structuring diverse willow
(genus Salix) communities. – Ecology 93: 138–150.
Siefert, A. et al. 2012. Functional beta-diversity patterns reveal
deterministic community assembly processes in eastern
North American trees. – Global Ecol. Biogeogr. 22: 682–691.
Smith, W. B. et al. 2009. Forest resources of the United
States, 2007. – USDA For. Serv. Gen. Tech. Rep. WO-78.
Spasojevic, M. J. and Suding, K. N. 2012. Inferring community
assembly mechanisms from functional diversity patterns:
the importance of multiple assembly processes. – J. Ecol.
Stephenson, N. 1990. Climatic control of vegetation distribution:
the role of the water balance. – Am. Nat. 135: 649–670.
Stephenson, N. L. 1998. Actual evapotranspiration and deﬁcit:
biologically meaningful correlates of vegetation distribution
across spatial scales. – J. Biogeogr. 25: 855–870.
Stevens, P. F. 2012. Angiosperm Phylogeny Website ver. 11.
Supplementary material (Appendix ECOG-00473 at
,www.oikosoﬃce.lu.se/appendix.). Appendix 1–6.