ArticlePDF Available

Abstract and Figures

Due to massive energetic investments in woody support structures, trees are subject to unique physiological, mechanical, and ecological pressures not experienced by herbaceous plants. Despite a wealth of studies exploring trait relationships across the entire plant kingdom, the dominant traits underpinning these unique aspects of tree form and function remain unclear. Here, by considering 18 functional traits, encompassing leaf, seed, bark, wood, crown, and root characteristics, we quantify the multidimensional relationships in tree trait expression. We find that nearly half of trait variation is captured by two axes: one reflecting leaf economics, the other reflecting tree size and competition for light. Yet these orthogonal axes reveal strong environmental convergence, exhibiting correlated responses to temperature, moisture, and elevation. By subsequently exploring multidimensional trait relationships, we show that the full dimensionality of trait space is captured by eight distinct clusters, each reflecting a unique aspect of tree form and function. Collectively, this work identifies a core set of traits needed to quantify global patterns in functional biodiversity, and it contributes to our fundamental understanding of the functioning of forests worldwide.
This content is subject to copyright. Terms and conditions apply.
Global relationships in tree functional traits
Daniel S. Maynard 1, Lalasia Bialic-Murphy 1, Constantin M. Zohner 1, Colin Averill1,
Johan van den Hoogen1, Haozhi Ma 1, Lidong Mo 1, Gabriel Reuben Smith 1,2, Alicia T. R. Acosta3,
Isabelle Aubin 4, Erika Berenguer 5,6, Coline C. F. Boonman 7, Jane A. Catford 8,
Bruno E. L. Cerabolini 9, Arildo S. Dias 10, Andrés González-Melo11, Peter Hietz 12, Christopher H. Lusk13,
Akira S. Mori 14, Ülo Niinemets15, Valério D. Pillar 16, Bruno X. Pinho 17,18, Julieta A. Rosell 19,
Frank M. Schurr20, Serge N. Sheremetev 21, Ana Carolina da Silva 22, Ênio Sosinski 23,
Peter M. van Bodegom 24, Evan Weiher 25, Gerhard Bönisch26, Jens Kattge 26,27 &
Thomas W. Crowther 1
Due to massive energetic investments in woody support structures, trees are subject to
unique physiological, mechanical, and ecological pressures not experienced by herbaceous
plants. Despite a wealth of studies exploring trait relationships across the entire plant
kingdom, the dominant traits underpinning these unique aspects of tree form and function
remain unclear. Here, by considering 18 functional traits, encompassing leaf, seed, bark,
wood, crown, and root characteristics, we quantify the multidimensional relationships in tree
trait expression. We nd that nearly half of trait variation is captured by two axes: one
reecting leaf economics, the other reecting tree size and competition for light. Yet these
orthogonal axes reveal strong environmental convergence, exhibiting correlated responses to
temperature, moisture, and elevation. By subsequently exploring multidimensional trait
relationships, we show that the full dimensionality of trait space is captured by eight distinct
clusters, each reecting a unique aspect of tree form and function. Collectively, this work
identies a core set of traits needed to quantify global patterns in functional biodiversity, and
it contributes to our fundamental understanding of the functioning of forests worldwide. OPEN
A full list of author afliations appears at the end of the paper.
NATURE COMMUNICATIONS | (2022) 13:3185 | | 1
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Trees play a fundamental role in forested ecosystems, driv-
ing carbon capture, nutrient cycling, and water dynamics,
with forest ecosystems supporting enormous terrestrial
biodiversity and being critically important for livelihoods
worldwide1. The types of trees that can survive in a given location
ultimately depend on their functional traits, which include all the
physiological and morphological features that determine how
they interact with, inuence, and respond to their environment2.
By governing treeswater, nutrient, and light economies26,
functional traits which elevate performance in one habitat typi-
cally reduce performance in others, leading to selection for spe-
cic traits across environments7. Genetic, morphological, and
biophysical constraints subsequently limit the range of traits that
a species can exhibit, leading to so-called ecological trade-offs
that shape speciesgeographic distributions8, coexistence
mechanisms9, and the provision of ecosystem services1,10. Iden-
tifying the key relationships that underpin global variation in tree
trait expression is needed for quantifying the functional biodi-
versity of forests, and will be critical for predicting how forest
diversity, composition, and function will respond to changing
environmental conditions11,12.
Despite a wealth of research into trait relationships across the
plant kingdom, our understanding of organismal-level trait
coordination in trees lags behind that of herbaceous species6,12,13.
Prior research has identied a key set of functional traits that
summarize the spectrum of form and function across the plant
kingdom, with leaf economics and plant size being the dominant
trait axes underpinning life-history strategies5,1416. Yet, because
such studies have traditionally focused only on a small handful of
traits estimated at the species leveloften heavily biased towards
leaf traits17the full dimensionality of trait space remains
unknown. This is particularly problematic for trees, which, due to
their size, longevity, ontogeny, and unique structural properties,
have distinct characteristics and face novel abiotic stressors
relative to herbaceous plants6,13,1821. Trait analyses which
include both woody and herbaceous plants typically omit critical
aspects of tree architecture (e.g. bark properties or crown size), as
these traits are either altogether absent in herbaceous plants or
because they are rarely measured on non-woody species. Thus,
our current understanding of dominant trait patterns in plants
fundamentally overlooks the massive energetic investments in
structures that are unique to large woody species21,22.
A challenge when measuring and quantifying forest functional
biodiversity is the enormous range of putative traits that can be
measured, in tandem with the relative difculty of measuring
many tree traits relative to herbaceous plants (e.g. root depth)2.
As with all plants, the functional biogeography of trees is known
to be partly governed by leaf economics13, which are critical for
light acquisition and water balance, as well as wood and root
traits, which are fundamental for structural stability and water
transport and nutrient acquisition6,23. Traits such as seed char-
acteristics and crown dimensions can be critical drivers of suc-
cessional trajectories, canopy positioning, and life-history
strategies across the landscape2426. Identifying the multi-
dimensional trait relationships that integrate these different tree
parts is urgently needed to create standardized, systematic metrics
of forest functional biodiversity across the globe12,27,28.
A key limitation when exploring functional trait relationships
is the sparsity inherent in most plant trait databases5,17,29,30. Such
restrictions limit traditional analyses to a small subset of traits
and species where there is complete coverage, often capturing
only a tiny fraction of known species (e.g. <1% of all plants5). To
overcome this limitation, studies increasingly use imputation
approaches to estimate species-level trait averages using phylo-
genetic and taxonomic information17,30,31. Yet, because these
approaches focus solely on species-level averages (i.e. gap-lling
of the species-by-trait matrix30), they are inherently biased
towards traits which are highly phylogenetically conserved and
which exhibit minimal environmental plasticity or intraspecic
variation29. To the extent that environmentally sensitive traits are
even included, their role in individual-level trait relationships can
be underestimated due to these species-level trait averages having
little ecological relevance32. To explore relationships between
phylogenetically conserved traits and environmentally plastic
traits, trait-imputation methods must consider evolutionary his-
tory along with the local environmental conditions32. Given the
enormous ontogenetic and environmental variation observed in
many important tree traitsincluding stem size, crown size, leaf
geometry, tree height, and root depth2325incorporating both
phylogenetic information and environmental information is
particularly important when exploring trait patterns in trees.
Here, we use a global trait database29 comprising nearly
500,000 trait measurements across more than 13,000 tree species
to explore relationships among 18 functional traits, reecting leaf
economics, wood structure, bark thickness, tree size and crown
size, seed size, and root depth (Fig. 1b). To overcome the data
sparsity and focus on relationships across all 18 traits, we develop
a series of non-parametric machine-learning models that provide
spatially-explicit estimates of trait expression of an individual tree
in a given location, as a function of its evolutionary history and
the local environmental conditions. Using the resulting dataset,
we ask: (1) Which traits and environmental variables best capture
overall variation in tree trait expression? and (2) What is the
dimensionality of trait space and the dominant multi-trait clus-
ters that capture the full breadth of tree form and function at the
global scale?
Our expectation was that traits underpinning the leaf-
economic spectrum would continue to emerge as a key driver
of trait relationships, primarily reecting the broad differences in
leaf morphology and physiology between angiosperms and
gymnosperms5. Yet, due to the wide range of traits being con-
sidered here, we expected the dimensionality of the trait space to
be more complex than when considering a small set of traits, or
when omitting traits integral to tree functioning (e.g. wood,
crown, or bark traits). Indeed, our analysis shows that the
dominant trait axes underpinning trait variation in trees closely
mirror those found across the plant kingdom5, with the rst axis
representing leaf-level resource economies, and the second axis
capturing whole-tree size and light competition. Yet these two
axes account for less than half of overall trait variation in trees,
suggesting more complex trait associations governing the full
variation in life-history strategies. By subsequently exploring
multidimensional relationships across all traits, we identify a
unique set of eight functional clusters that reects the full breadth
of tree form and function, and which can aid in trait selection for
global studies of forest functional biodiversity. Collectively, this
work identies emergent constraints on tree functional biogeo-
graphy, and sheds light on the core set of traits needed to quantify
and study forest functional diversity worldwide.
Trait models. Our analysis included 491,001 unique trait mea-
surements across 18 traits, encompassing 13,189 tree species from
2313 genera, reecting ~21% of all known tree species33 (Fig. 1).
Traits were measured at 8683 locations across the globe and 373
distinct eco-regions (Supplementary Tables 1, 2), with georefer-
enced measurements capturing 15% of known tree species in
Eurasia, 13% in South America, 9% in Oceania, and 6% in North
America and Africa33. The raw data covered 22% of all trait-by-
species combinations (Fig. 1b, Supplementary Fig. 2), nearly
identical to other large-scale trait analyses across the entire plant
2NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
kingdom5,17,30. Yet there was considerable variation in coverage
across traits, with traits such as specic leaf area and leaf nitrogen
measured on more than 60% of all species, versus traits such as
crown diameter and conduit diameter, which captured fewer than
5% of species (Fig. 1b, Supplementary Fig. 2). Across all species,
423 had more than 10 unique traits measured, and two species
(Picea abies and Pinus sylvestris) had measurements for all 18
traits. In general, there was highly consistent coverage across
taxonomic orders and traits (Supplementary Fig. 1), with gym-
nosperms being slightly overrepresented (comprising 3.1 ± 6.8%
of measurements in the database versus ~1% of all known tree
species34,35, Fig. 1a), in part reecting the wider geographic range
of many gymnosperms relative to angiosperms36.
To explore relationships in functional traits at the individual
level, we used random-forest machine-learning models to
estimate missing trait values for each individual tree as a function
of its environment and phylogenetic history. We also conducted a
second set of analyses where trait expression was estimated using
phylogenetic information only, which allowed us to include
additional non-georeferenced data (Fig. 1), while also quantifying
the relative contribution of environmental information on trait
expression (Supplementary Fig. 6). Following standard
approaches5,15,29,30, all traits were log-transformed and standar-
dized to allow for statistically robust comparisons. Environmental
predictors included ten variables encompassing climate3740,
soil41, topographic42, and geological43 features. Phylogenetic
history was incorporated via the rst ten phylogenetic
eigenvectors44,45 (see Methods). By including environmental
information alongside phylogenetic information, this approach
not only allowed us to impute species-level traits which have
strong phylogenetic signals and weak environmental signals, as is
traditionally done17,30 but also to robustly estimate traits which
have a weak phylogenetic signal and are instead strongly sensitive
to environmental conditions. Moreover, being a non-parametric
approach, the random forest makes no a priori assumptions
about how trait expression varies across phylogenetic groups or
Across all 18 traits, the best-tting models explained 54 ± 14%
of out-of-t trait variation (VEcv, see Methods), ranging from
26% for stem diameter to 76% of the variation in leaf area
(Supplementary Figs. 6, 7). This accuracy was quantied using
buffered leave-one-out cross-validation to account for spatial and
Fig. 1 Overview of the 18 functional traits. a The unique geographic locations (n=8683) where tree functional traits were recorded. The size of the circles
denotes the relative number of unique traits (out of 18 possible) that were measured at each location, regardless of species identity. bSummary statistics
for the 18 traits considered here (see Supplementary Table 13, Supplementary Figs. 1, 2 for additional information). The analysis included 491,001 trait
measurements, encompassing 13,189 unique tree species and 2313 unique genera.
NATURE COMMUNICATIONS | (2022) 13:3185 | | 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved
phylogenetic autocorrelation46, and thus serves as a conservative
lower bound for species which are phylogenetically and
environmentally distinct from the observations47. There was no
signicant relationship between out-of-t cross-validation accu-
racy and sample size (R2=0.06, p=0.33), highlighting the
relatively broad taxonomic coverage for each trait (Fig. 1,
Supplementary Fig. 1).
Environmental variables and phylogenetic information had
approximately equal explanatory power (relative importance of
0.51 vs 0.49 for environment vs. phylogeny), albeit with
substantial variation across traits (Supplementary Fig. 9). The
inclusion of environmental variables increased the explanatory
power of the models by 35%, on average (Supplementary Fig. 6),
with crown diameter, crown height, leaf density, and stem
diameter exhibiting the largest relative increases (54%, 45%, 73%,
and 26%, respectively), mirroring the fact that these traits have
comparatively low phylogenetic signal relative to other traits
(assessed via Pagelsλon the raw data, Fig. 4c). Seed dry mass was
the only trait with a substantial increase in accuracy using the
phylogeny-only model (25% improvement; Supplementary Fig. 6),
reecting the fact that seed dry mass had the strongest
phylogenetic signal of all traits (Fig. 4c), and also because this
trait has a substantial amount of additional non-georeferenced
data that was included in the phylogeny-only models (Fig. 1b).
Wood density was the only trait with nearly identical predictive
power whether or not environmental information was included,
whereas all other traits exhibited signicantly reduced accuracy
when environmental information was excluded (Supplementary
Fig. 6).
Relationships in tree trait expression. Using the resulting trait
models, we imputed missing trait values for every tree with at
least one georeferenced trait measurement. For all traits except
seed dry mass, we used the random-forest models accounting for
environmental and phylogenetic information; for seed dry mass,
we used the phylogeny-only model to estimate expression due to
its substantially higher data availability and out-of-t accuracy.
For tree height, stem diameter, crown height, crown width, and
root depth, we used quantile random forest48 to estimate the
upper 90th percentile value for each species in its given location,
thereby minimizing ontogenetic variation across a trees lifetime
(see Methods). We used the resulting trait data to explore the
dominant drivers of trait variation using species-weighted prin-
cipal component analysis, accounting for an unequal number of
observations across species.
When considering all traits simultaneously, the rst two axes of
the resulting principal components (PC) capture 41% of the
variation in overall trait expression (Fig. 2a; Supplementary
Fig. 10; Supplementary Table 5). The rst trait axis correlates
most strongly with leaf thickness, specic leaf area, and leaf
nitrogen (PC loadings of L =0.77, 0.74, and 0.73, respectively).
By capturing key aspects of the leaf-economic spectrum14, these
traits reect various physiological controls on leaf-level resource
processing, tissue turnover and photosynthetic rates49. Thick
leaves with low specic leaf area (SLA) can help minimize
desiccation, frost damage, and nutrient limitation, but at the cost
of reduced photosynthetic potential due to primary investment in
structural resistance50. Accordingly, leaf nitrogena crucial
component of Rubisco for photosynthesis51trades off strongly
with leaf thickness. This rst axis thus captures the core
distinction between acquisitive(fast) and conservative(slow)
life-history strategies across the plant kingdom7,52,reecting an
organismal-level trade-off between the high photosynthetic
potential in optimal conditions versus abiotic tolerance in
suboptimal conditions. Nevertheless, leaf densitywhich is
related to SLA and is a key feature of the leaf-economic
spectrumloads relatively weakly on this rst trait axis compared
to other leaf traits (L =0.28 for axis 1, vs 0.20 for axis 2;
Supplementary Table 5), highlighting important aspects of leaf
structure that are not captured by this dominant trait axis53.
The second trait axis correlates most strongly with maximum
tree height (PC loading of L =0.77), crown height, (L =0.75),
and crown diameter (L =0.88), highlighting the overarching
importance of competition for light and canopy position in
forests7(Fig. 2a; Supplementary Fig. 10; Supplementary Table 5).
Large trees and large crowns are critical for light access and for
maximizing light interception down through the canopy54.
Nevertheless, tall trees with deep crowns also experience greater
susceptibility to disturbance and mechanical damage, primarily
due to wind and weight25. Because of the massive carbon and
nutrient costs required to create large woody structures55,56,
larger trees are less viable in nutrient-limited or colder climates57,
and in exposed areas with high winds or extreme weather
events58. This second axis thus reects a fundamental biotic/
abiotic trade-off related to overall tree size, which is largely
orthogonal to leaf-level nutrient-use and photosynthetic capacity.
Despite substantial differences in wood and leaf structures
between angiosperms and gymnosperms (e.g. vessels vs. trac-
heids), the two main relationships hold within, as well as across,
angiosperms and gymnosperms (Fig. 2b, c; Supplementary
Figs. 11, 12). Indeed, angiosperms and gymnosperms are subject
to the same physical, mechanical, and chemical processes that
determine the ability to withstand various biotic and abiotic
Collectively, these two primary trait axes capture two dominant
ecological trade-offs that underpin tree survival in any given
environment: (1) the ability to maximize leaf photosynthetic
activity, at the cost of increased risk of leaf desiccation, and (2)
the ability to compete for space and maximize light interception,
at the cost of increased susceptibility to mechanical damage. By
capturing two aspects of conservative-acquisitive life-history
strategies, these two relationships closely mirror those seen when
considering herbaceous species alongside woody species5,17.
However, in line with our expectations, these two axes capture
only ~40% of the variation in trait space, versus nearly ~75% of
variation when considering only six traits across the entire plant
kingdom5. Here, the rst seven PC axes are needed to account for
75% of the variation across all 18 traits (Supplementary Table 5).
Thus, while this analysis supports the universality of these two
primary PC axes, it also demonstrates that the majority of trait
variation in trees is unexplained by these two dimensions. As
such, quantifying the full dimensionality of trait space by
exploring multidimensional trait clusters is needed to better
capture the wide breadth of tree form and function.
Environmental predictors of trait relationships. To examine
how environmental variation shapes trait expression across the
globe, we next quantied the relationships between environ-
mental conditions and the dominant trait axes. Using Shapley
values60, we partitioned the relative inuence of each environ-
mental variable on the PC trait axes, controlling for all other
variables in the model (see Methods).
In line with previous analysis across the plant kingdom61,
temperature variables were the strongest drivers of trait relation-
ships (Fig. 3, Supplementary Figs. 17, 18), with annual temperature
having the strongest inuence both on leaf-economic traits (PC
axis 1, Fig. 3c) and on tree-size traits (PC axis 2, Fig. 3d). Leaves
face increased frost risk and reduced photosynthetic potential in
colder conditions, such that ecological selection should favour
thick leaves with low SLA over thin leaves with high SLA and high
4NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
nutrient-use49. Trees in warm environments are more likely to
experience strong biotic interactions, which should increase
evolutionary and ecological selection pressures over time62,63,
favouring tall species with large crowns that have high competitive
ability and efcient light acquisition strategies. Annual temperature
thus predominantly reects the transition from gymnosperm- to
angiosperm-dominated ecosystems, with this inection point
occurring at ~15°C for both axes, demonstrating strong environ-
mental convergence between the dominant axes of trait variation.
Beyond annual temperature, each trait axis demonstrated
different relationships with climate, soil, and topographic
variables (Fig. 3a, b, Supplementary Figs. 17, 18). Percent sand
Fig. 2 The dominant trait axes and relationships. Shown are the rst two principal component axes capturing trait relationships across the 18 functional
traits. aAll tree species (n=30,146 observations), bangiosperms only (n=24,658), and cgymnosperms only (n=5498). In athe three variables that
load most strongly on each axis are shown in dark black lines, with the remaining variables shown in light grey. These same six variables are highlightedin
band cillustrating how the same relationships extend to angiosperms and gymnosperms (see Supplementary Figs. 1012 for the full PCAs with all traits
visible, and Supplementary Table 5 for the PC loadings).
Fig. 3 The relationship between environmental variables and trait axes. a,bThe relative inuence of the environmental variables on the two dominant
PC axes. The ten variables are sorted by overall variable importance in the models (see Methods). Yellow points are observations which have high values
of that environmental variable; blue values are the lowest. Points to the right of zero indicate a positive inuence on the PC axis; points to the left indicate a
negative inuence (see also Supplementary Figs. 17, 18). chThe relationships between environmental variables and PC axis values for the three variables
in awith the strongest inuence. Values above zero show a positive inuence on PC axis values; values less than zero indicate a negative inuence.
NATURE COMMUNICATIONS | (2022) 13:3185 | | 5
Content courtesy of Springer Nature, terms of use apply. Rights reserved
content had the second-highest inuence on the rst trait axis
(Fig. 3e), supporting patterns seen across the entire plant
kingdom17. Sand content is a strong proxy for soil moisture
and soil-available nutrients such as phosphorous, and is therefore
closely tied to leaf photosynthetic rates64. In contrast to previous
work, however, we nd that soil characteristics have correspond-
ingly little effect on the second axis of trait variation (Fig. 3b;
Supplementary Fig. 18). Instead, precipitation was the second
strongest driver of tree height and crown size (Fig. 3f), with large
trees with large crowns becoming consistently more frequent with
increasing precipitation. These results highlight that, despite the
primary importance of temperature, the main climate stressors to
trees (e.g. xylem cavitation and embolism, re regimes, and leaf
desiccation) typically arise via interactions between temperature,
soil nutrients, and water availability.
For both axes, elevation was the third strongest driver of trait
values (Fig. 3g, h), highlighting a critical component of tree
functional biogeography that extends beyond climate and soil. Yet
the effects of elevation on trait expression differed somewhat
across the two axes. For the rst axis related to leaf-economic
traits, there is little inuence at low elevations, followed by a
sharp transition at ~2000 m towards gymnosperm-dominated
species with thick leaves, low SLA, and low leaf N. For the second
trait axis related to tree size, elevation instead has a strong
positive inuence on tree height and crown size at low elevations,
which becomes increasingly less inuential past ~500 m. Such
results partly reect the transition from angiosperm to
gymnosperm-dominated stands at higher elevations (blue vs.
red points, Fig. 3g, h), and potentially the role of environmentally
mediated intraspecic variation in traits such as tree height65,66.
These results demonstrate close alignment of the dominant
trait PC axes across biogeographic regions. Despite the orthogon-
ality of these axes in trait species, environmental conditions place
similar constraints on both trait axes, particularly at the
environmental extremes (e.g. warm, moist, low elevation vs. cold,
dry, high elevation), leading to convergence of the dominant trait
axes across environmental gradients.
Trait clusters at the global scale. To better explore the multi-
dimensional nature of trait relationships that are not fully covered
by the dominant two axes, we subsequently identied groups of
traits that form tightly coupled clusters and which reect distinct
aspects of tree form and function.
Our results show that these 18 traits can be grouped into eight
trait clusters, each of which reects a unique aspect of morphology,
physiology, or ecology (Fig. 4a, Supplementary Fig. 23). The largest
trait cluster (Fig. 4a, pink cluster) demonstrates wood/leaf
integration of moisture regulation and photosynthetic activity via
the inclusion of leaf area, stem conduit diameter, stomatal
conductance, and leaf V
(the maximum rate of carboxylation).
Distinct from this cluster are the three traits loading most strongly
on PC axis 1 (SLA, leaf thickness and leaf N; Fig. 4a, yellow),
highlighting complementary aspects of the leaf-economic spectrum
indicative of acquisitive vs. conservative resource use15. The role of
leaf K and P in leaf nutrient economies are well established7,67,and
yet these traits form a distinct cluster from the other leaf-economic
traits (Fig. 4a, light blue) due to their relatively high correlation
with tree height and crown size, particularly for leaf K, which loads
almost equally on both trait axes (Fig. 4b, Supplementary Table 5).
Tree height and crown size form their own distinct cluster
(Fig. 4a, dark green), further supporting the inference that these
traits reect key aspects of tree form and function independent of
the leaf-economic spectrum. Yet leaf area, despite being part of
the cluster reecting moisture regulation and photosynthetic
activity, loads almost equally on PC axes 1 and 2 (Fig. 4b,
Supplementary Table 5), highlighting that it serves as an
intermediary between the two key aspects of tree size and leaf
economics. It is a critical driver of moisture regulation and
photosynthetic capacity, while also playing an important role in
the light acquisition, leaf-turnover time, and competitive
There are two additional two-trait clusters, both of which load
relatively poorly on the two primary PC axes: (1) stem diameter
and bark thickness (Fig. 4, dark blue), and (2) wood and leaf
density (Fig. 4, light green). Bark thickness increases with tree size
not only as a result of bark accumulation as trees age, but also due
to the functional/metabolic needs of the plant69,70. From an
ecological perspective, thick bark can be critical for defense
against re and pest damage (mainly a thick outer bark region),
for storage and photosynthate transportation needs (mainly a
thick inner bark region)71,72. Yet such relationships are strongly
ecosystem-dependent, with tree size emerging as the dominant
driver at the global scale70. In contrast, wood density and leaf
density are strongly linked to slow/fast life-history strategies,
where denser plant parts reduce growth rate and water
transport6,15 but protect against pest damage, desiccation, and
mechanical breakage6,50,56. As such, leaf density captures
fundamentally unique aspects of leaf form and function relative
to other leaf traits such as SLA53 (Fig. 4b, Supplementary
Table 5), and our results support the inference that these translate
into fundamentally different ecological strategies73. Collectively,
these two-trait clusters each demonstrate unique and comple-
mentary mechanisms that insulate trees against various dis-
turbances and extreme weather events, but at the cost of reduced
growth, competitive ability, and productivity under optimal
conditions (see Supplementary Notes).
Lastly, two traits each comprise their own unique cluster: root
depth and seed dry mass (Fig. 4a, purple and orange,
respectively). Root growth is subject to a range of belowground
processes (e.g. root herbivory, depth to bedrock), and our results
conrm previous work demonstrating a clear disconnect between
aboveground and belowground traits23,74,75. Root depth accord-
ingly has a relatively weak phylogenetic signal (λ=0.44, Fig. 4c)
but a strong environmental signal (Supplementary Figs. 6, 9),
reecting distinct belowground constraints on trait expression23.
In contrast, seed dry mass exhibits the strongest phylogenetic
signal (λ=0.98, Fig. 4c) and weakest environmental signal of any
trait (Supplementary Figs. 6, 9), and it accordingly was the only
trait where the phylogeny-only model performed substantially
better (Supplementary Fig. 6). In line with previous work, seed
dry mass has moderate correlations with various other traits
underpinning leaf economics and tree size5,28 (e.g. ρ=0.28,
0.22, and 0.22 for tree height, leaf K, and leaf density, using the
raw data), yet it exhibits relatively weak correlation with most
other traits, placing it in a distinct functional cluster. Reproduc-
tive traits are subject to unique evolutionary pressures26,
indicative of different seed dispersal vectors (wind, water,
animals) and various ecological stressors that uniquely affect
seed viability and germination26. The emergence of root depth
and seed dry mass as solo functional clusters thus supports the
previous inference that belowground traits74 and reproductive
traits26 reect distinct aspects of tree form and function not fully
captured by leaf or wood trait spectrums.
This work provides a baseline understanding of tree form and
function, aimed at informing future research into the functional
biodiversity of forests. Our lack of understanding of the trait
relationships unique to woody species is partly due to the enor-
mous number of putative traits that can be measured on trees7,29.
6NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
To help address these challenges, the eight distinct trait clusters
identied here (Fig. 4) can help inform future research into forest
functional biodiversity. First, this work indicates traits which are
largely redundant at the global scale (e.g. SLA and leaf N), versus
those that occupy clear distinct roles (seed dry mass and root
depth). Second, by quantifying the relative importance of envir-
onmental conditions vs. phylogenetic history as drivers of trait
expression (Fig. 4c, Supplementary Figs. 6, 9), these results can
aid in the selection of representative traits from each cluster,
depending on the intended question or application (e.g. by
selecting environmentally sensitive traits such as crown size, vs.
more phylogenetically conserved traits such as maximum tree
height). Such insight can help with the adoption of a standardized
set of tree traits that allow for consistent quantication and
comparison of forest biodiversity on a global scale.
The 18 physiological and morphological traits used here were
selected in part due to their representation in prior trait analyses,
their functional uniqueness, their relevance to tree structure and
architecture, and the data quality. Thus, as with all functional trait
analyses, there are additional traits and metrics not considered
here which may capture complementary aspects of tree form and
function. For example, various allometric relationships (e.g. the
ratio of tree height to stem diameter) may better capture energetic
allocation in trees than do physiological traits76 by reecting
unique aspects of tree morphology (see Supplementary Figs. 24,
25 for analysis exploring allometric ratios). Alternatively, below-
ground traits such as root architecture and chemistry remain
particularly underrepresented for trees, despite capturing poten-
tially unique aspects of tree form and function74,75. Increasing the
representation of such traits in trait databases is an important
next step, and critical for quantifying the multi-functionality of
forest biodiversity.
Here, by focusing on trees, we have signicantly higher species-
level representation compared to kingdom-wide analyses,
encompassing more than 20% of known tree species and 22% of
all trait-by-species combinations (Supplementary Fig. 2), com-
parable to other large-scale analyses exploring trait relationships
across the entire plant kingdom5,17,30. However, because we
explore patterns obtained using imputed datawhich can
introduce issues of circularity and imputation biaswe con-
ducted a series of sensitivity analyses to test the robustness of our
ndings to various modelling assumptions. Most importantly,
when using the raw (un-imputed) data to estimate pairwise
correlations, we see that the PCA results, functional clusters, and
environmental relationships are nearly identical to those obtained
using the imputed data (Supplementary Fig. 2023). Additionally,
our results are likewise insensitive to the selection of phylogeny-
only vs. phylogeny +environment models (Supplementary
Fig. 13), to the use of all 52,255 tree species in the reference
phylogeny (Supplementary Fig. 14), to various levels of miss-
ingness in the data (Supplementary Fig. 15), and to the number of
predictor variables in the models (Supplementary Fig. 16).
Indeed, the value of the PCA and clustering approaches used here
is that they rely only on accurate estimates of pairwise correla-
tions, which have shown to be highly robust to data sparsity5,30.
Moreover, by focusing only on correlative analyses, we can
directly validate these results using the raw data (Supplementary
Figs. 2023).
Nevertheless, an ongoing challenge in plant trait analyses is to
improve the taxonomic and spatial representation of trait
Fig. 4 Trait correlations and functional clusters. a Trait clusters with high average intra-group correlation. The upper triangle gives the species-weighted
correlations incorporating intraspecic variation. The lower triangle gives the corresponding correlations among phylogenetic independent contrasts, which
adjusts for pseudo-replication due to the non-independence of closely related species. The size of the circle denotes the relative strength of the correlation,
with solid circles denoting positive correlations and open circles denoting negative correlations (see Supplementary Fig. 19 for the numeric values). bPC
loadings for each trait and each of the rst two principal component axes, illustrating which functional trait clusters align most strongly with the dominant
axes of trait variation (see Supplementary Table 5 for the full set of PC loadings). cThe species-level phylogenetic signal of each trait (Pagelsλ), calculated
using only the raw trait values.
NATURE COMMUNICATIONS | (2022) 13:3185 | | 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved
measurements29. We had generally broad spatial coverage (Sup-
plementary Tables 1, 2; Fig. 1), with notable exceptions being the
interior of Africa and northeast Asia, where trait data is lacking in
general29. Similarly, we had largely consistent representation
across taxonomic orders (Fig. 1, Supplementary Fig. 1), albeit
with variation across traits and orders, with species in the order
Pinales, for example, being consistently overrepresented (com-
prising 4% of observations vs ~1% of known tree species; Sup-
plementary Fig. 1), versus those in the orders Asterales and
Solanales being slightly underrepresented (1% of observations vs
~2% of known tree species). By using buffered leave-one-out
cross-validation46, our accuracy estimates present a conservative
lower bound for such species47 (Supplementary Fig. 6). None-
theless, caution should still be exercised when using trait-
imputation approaches to make inferences of trait values for a
specic tree in a given location, particularly on the unlogged scale.
Nearly all traits exhibit skewed log-normal distributions (where
the sample variance is proportional to the mean), and so traits
must be log-transformed to allow for statistically valid
comparisons5,15,29,30. As such, estimates on the raw (linear) scale
can be subject to unavoidably high variation, regardless of model
accuracy (here, 29 ± 19% median relative absolute error on the
unlogged scale, ranging from 12% for wood density to 86% for
seed dry mass). Thus, although our approach yields robust
insights into global patterns that hold when making comparative
analyses of aggregate trends across species5,17 (Supplementary
Fig. 8), obtaining highly precise estimates of trait expression for a
given tree in a given location remains a fundamental challenge in
functional ecology.
Our analysis differs from previous approaches5,17,30 in that we
use non-parametric machine-learning models to estimate trait
expression as a function of phylogenetic history and environ-
mental variables. This allows us to model traits regardless of their
degree of phylogenetic signal or relationships to abiotic condi-
tions (Fig. 4c, Supplementary Figs. 6, 9). Yet there are likely
additional abiotic and biotic factors not included here which are
important drivers of trait expression. For example, forest age and
successional stage are key drivers of tree morphology and stand
structure77; and forest management practices, disturbance his-
tory, human activity, and native/introduced status can all play
important roles in tree trait expression78 (see Supplementary
Notes). Local environmental conditions (e.g. sun vs. shade,
microsite soil characteristics) are also critical drivers of individual
tree morphology and physiology, and landscape-level climate
variables like those used here may underestimate the extent of
these local abiotic controls79. By including environmental varia-
tion alongside phylogenetic information, our approach should
better incorporate these non-taxonomic drivers of trait expression
than approaches which do not allow for intraspecic variation.
Yet including high-resolution microsite data and stand-level
information in existing trait databases will be critical for gaining a
more nuanced understanding of the drivers of tree trait expres-
sion. Here, by quantifying the full dimensionality of trait space,
this work can serve as a baseline for such research, helping to
identify the dominant traits that underpin the functional biodi-
versity of forests.
Collectively, this work reveals key relationships and trait
clusters governing tree form and function worldwide. We show
that tree functional traits predominantly reect two major func-
tional axes: one representing leaf-level photosynthetic capacity
and resource economies, and the other representing competition
for light via tree and crown size. Mirroring patterns seen across
the entire plant kingdom, these patterns capture an ecological
gradient from conservative strategies under suboptimal environ-
ments (cold, dry, high elevation), to acquisitive strategies asso-
ciated with light competition in high-resource environments
(warm, high soil quality, low elevation). However, these two axes
capture less than half of the overall variation in tree form and
function. By subsequently exploring multidimensional relation-
ships across all traits, we identify a unique set of eight functional
clusters that reects the full breadth of tree form and function. In
doing so, these results elucidate key constraints on functional trait
relationships in trees, contributing to our fundamental under-
standing of the controls on the function, distribution, and com-
position of forest communities. By identifying a core set of traits
that reect the broad variety of ecological life-history strategies in
trees, this work can inform future trait-based research into the
functional biodiversity of the global forest system.
Trait information. Trait data were obtained from the TRY plant trait database29 v.
5.0. Data were cleaned by converting all traits to standardized units and by
matching species names to The Plant List (TPL) database v1.1 (http://www., accessed June 2020) using the Taxonstand package in R v3.6.080.
Synonyms were replaced with accepted names, when available. The phylogenetic
tree was taken from the seed plant phylogeny of Smith & Brown81, and species
names were likewise cleaned and harmonized using the TPL database. To limit our
analysis to tree functional traits only, we used the BGCI GlobalTreeSearch database
v1.335,82, containing a comprehensive list of ca. 60,000 tree species compiled and
harmonized from across 500 sources. The BGCI database uses TPL for much of its
taxonomic identication, but to ensure consistency among all sources we used the
same name harmonization pipeline as with TRY and the seed plant phylogeny. We
constrained the set of traits and the phylogenetic tree to those species that could be
matched to the BGCI database (n=52,255 species matched, excluding monocots),
and we likewise trimmed the phylogenetic tree to the set of species matched in TRY
For comparison with previous work on trait trade-offs, we focused on
physiological and morphological traits directly measured at the individual level,
rather than derived traits such as those related to tree allometry (although see
Supplementary Figs. 24, 25 and Supplementary Notes). The core set of 18 traits was
selected from the TRY database rst based on their use in prior global analyses5
(leaf area, specic leaf area, seed dry mass, maximum tree height, leaf nitrogen,
wood density), their importance in the leaf economics spectrum14,15,20,83 (leaf
thickness, density, Vcmax, phosphorous, potassium, and stomatal conductance),
their role in tree water transport and access23,84 (root depth, stem conduit
diameter), and nally, those that are integral components of tree structure6(stem
diameter, bark thickness, crown height and diameter). This resulted in a primary
set of 18 traits for use in the main analysis (Supplementary Table 3, Supplementary
Data 1). For each trait, we selected sub-categories (as specied by TRY) that
denoted comparable measurements and reected uniform assay conditions (e.g.
measured at 25 °C) (Supplementary Table 3). We also obtained an additional
12 traits from the TRY to improve our ability to predict missing trait values via
incorporating trait covariation (see model details, below). However, these were not
used in the main analysis because they are either auto-correlated with one of the
focal traits by denition (e.g. leaf width, leaf length, leaf C/N, leaf N/P, leaf
moisture), they encompassed fewer than 150 species (root/shoot ratio, root N per
mass, root C/N, and wood N per mass), or because they can vary substantially
across measurement protocols or assay conditions but had insufcient metadata
(i.e. lack of incubation temperature) needed for standardization (leaf J
, leaf
respiration rate, leaf chlorophyll). Trait values were converted to common units
where necessary (e.g. mm to cm), log-transformed, and normalized to have a mean
of zero and a standard deviation of one, following standard approaches5,15,29,30.
Trait-imputation models. In order to consider multidimensional trait relation-
ships across all species and all traits, we used a machine-learning model to estimate
the missing trait values for each species in each location. Specically, we used
random-forest (RF) models to estimate trait values for each georeferenced species
using the ranger package in R85, with trait expression modelled as a function of
each trees environment and phylogenetic history. The random-forest algorithm is
determined by a specied set of hyper-parameters, which govern the splitting
rules, variable selection, and stopping rules when building each tree in the
forest86,87. Preliminary investigation showed that a wide variation in these hyper-
parameters (e.g. using 1 versus 25 minimum observations per node) led to negli-
gible improvements in model accuracy relative to the default parameters (i.e. <2%
increase in out-of-bag R2). Thus, to minimize the risk of overtting due to a large
number of traits and models being considered here, we used the default ranger
hyper-parameters for all traits (500 trees per forest; sampling with replacement; the
number of variables per split equal to the square root of the number of predictors; a
minimum of 5 observations per node; and the split rule determined by maximal
variance). In order to minimize the inuence of data-recording errors or unit
mismatches in the dataset, trait values which occurred outside of the bulk of the
trait distribution were investigated as outliers. Those which could not be externally
veried and which were biologically unreasonable were removed (e.g. stem dia-
meters > 15 m). When modeling tree height, crown size, and root depth, we only
8NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
considered observations with height > 5 m, stem diameter > 10 cm, ro ot depth >
25 cm, and crown height and width > 1 m88, thereby ensuring that our analysis
focused on adult trees rather than saplings or woody shrubs. We subsequently
implemented quantile random forest48,89 to estimate the upper 90th percentile trait
value for maximum stem diameter, crown height and width, and root depth. In all
other cases the imputed traits represent the mean predicted value across the ran-
dom forest.
Environmental covariates considered for use in the models included 50
variables encompassing a range of climate3739,soil
41, topographic42, and
geological43 variables (Supplementary Table 4). We omitted variables that directly
measure plant community composition or biotic factors (e.g. NDVI or % forest
cover) so as to ensure the resulting geographic layers solely encompassed non-
biotic factors. Layers were sampled from a previously prepared global composite90.
Briey, all covariate map layers were resampled and reprojected to a unied pixel
grid in EPSG:4326 (WGS84) at 30 arcsec resolution (~1 km2at the equator). Layers
with a higher original pixel resolution were downsampled using a mean
aggregation method; layers with a lower original resolution were resampled using
simple upsampling (that is, without interpolation) to align with the higher
resolution grid. The set of environmental variables for each trait measurement was
obtained by sampling this composite image at each unique latitude and longitude
value given in the TRY database.
Phylogenetic information was incorporated in the form of phylogenetic
eigenvectors31,32,44,45.Werst calculated the pairwise cophenetic phylogenetic
distance matrix across all 54,153 tree species that could be matched to both the
BGCI tree list and the plant phylogeny. This matrix was then double-centred by
rows and columns44,91, and the eigenvectors were sorted by percent variation
explained across the phylogeny, with the rst eigenvector accounting for the
majority of variation. The rst 50 orthogonal eigenvectors were extracted from this
matrix for consideration as continuous predictors in the random-forest models.
To improve model parsimony and minimize overtting, we used a sequential
variable selection approach, whereby we selected k=3, 5, 10, 15, 25, 50
phylogenetic and environmental variables (each), and t the full set of models. The
kphylogenetic eigenvectors for each step were selected by taking the rst 1,...,k
eigenvectors, sorted based on percent variation explained across the phylogeny.
The kenvironmental variables were selected via clustering the full set of 50
variables into kgroups and selecting a representative variable from each group,
thereby minimizing correlation. Doing so revealed that out-of-t model accuracy
saturates at approximately k=10 variables (Supplementary Fig. 5). We therefore
used 10 phylogenetic eigenvectors and 10 environmental variables in the nal
models to improve model parsimony and minimize overtting; however, the results
are unchanged regardless of the number of covariates (Supplementary Fig. 6).
To leverage trait covariation among the disparate observations, we used a two-
step algorithm to improve predictive power and imputation accuracy31,92. Using
the general approach of Stekhoven & Bühlmann (2012), we rst implemented a
random forest on all traits for all observations. We then used these initial models to
predict the full set of trait values for each observation (including the 12 ancillary
traits not included in the focal analysis, Supplementary Table 3). We then ret the
random-forest models for each trait, using the full set of predicted traits (apart
from the focal traits) as covariates. For the nal analysis, observed traits were used
in place of imputed traits, when available, with the exception of maximum tree
height, stem diameter, root depth, and crown size, where the upper 90th percentile
trait values were used. Variable importance in the random-forest models
(Supplementary Fig. 9) was calculated using the permutationmetric, reecting
the variance in responses across predictors85.
By incorporating using both phylogenetic and environmental variables in the
random-forest models, our approach makes no assumptions about which traits are
more strongly governed by phylogeny versus environmental conditions. It thus
provides a non-parametric alternative to taxonomic-based imputation methods30,
allowing for traits with high phylogenetic signal and low environmental
relationship, as well as those with low phylogenetic signal and high environmental
relationship. Nevertheless, because some traits had substantial amounts of non-
georeferenced information available, we conducted a second set of models using
only the phylogenetic eigenvectors as predictors. For each trait, the nal model
used to impute missing trait values was the one with the higher predictive accuracy
(either phylogeny-only, or phylogeny +environmental variables). Note that this
amounted to us using the phylogeny-only model for seed dry mass only, and the
combined model for all other traits (Supplementary Fig. 6).
Model performance. Model performance was quantied using buffered leave-one-
out cross-validation46. To avoid overestimating the out-of-t accuracy, we rst
estimated the range of spatial and phylogenetic autocorrelation in the raw data,
following the approach recommended in Roberts et al. (2017). Specically, before
running any trait-imputation models, we t a simple linear regression model to the
raw data, where trait expression was modelled as a linear function of phylogenetic
and environmental variables. We then assessed spatial autocorrelation of the
residuals via Morans I plots using the ncf package in R, which displays the value of
spatial autocorrelation (ranging from 1 to 1) as a function of distance93.We
likewise assessed residual phylogenetic autocorrelation across taxonomic ranks
(genus, family, order, group), using the the ape package in R. In general, spatial
autocorrelation was low (I <0.10) (Supplementary Fig. 3), with the exception of leaf
phosphorous, which exhibited slight autocorrelation up to ~250 km. The residual
phylogenetic correlation was likewise low, and generally only observable at the
genus level, apart from crown width and height and conduit diameter, which
exhibited residual autocorrelation up to the family level (Supplementary Fig. 4).
Thus, to be conservative, for all traits except crown size and conduit diameter, we
used a genus-level spatial buffer of 250 km to exclude test/training data; and for
crown size and conduit diameter we used a family-level buffer at 250 km.
To implement the cross-validation accuracy assessment, we rst randomly
selected a focal species, with the out-of-t test data containing all observations for
that species for the focal trait. To construct the corresponding training data, we
excluded all observations of the same genus (or family) that fell within a 250 km
spatial buffer of any of the training points for that species. The random-forest
models were then t using the buffered training data, and used to predict the trait
values for the omitted species46. This procedure was repeated for each unique
species for each trait, up to 1000 times, with a randomly sampled focal species
selected at each iteration. Note that this approach is known to underestimate the
actual accuracy47, and so the R2values should thus be seen as conservative lower
Following the recommendation of Li (2017), model accuracy was calculated via the
cross-validated coefcient of determination relative to the one-to-one line (termed "VEcv",
Li 2017), which provides a normalized version of the mean-squared-error (MSE) that
allows for comparisons across data types and units. Specically, this value is calculated as:
R2VECV ¼1ðypred
yÞ2¼1SSE=TSS ¼1MSE=^
the summation is taken across the species, and the predicted values are estimated out-of-t
using the buffered cross-validation procedure outlined above. Importantly, this metric is
not the same as a regression-based goodness-of-t, as it is instead calculated by direct
comparison of observed vs. out-of-t predicted values94.
Principal component analysis. To identify the dominant axes of variation across
all 18 traits, we use species-weighted principal component analysis (PCA). This was
conducted on the full set of imputed traits using the aroma.light package in R, with
observed values used in place of imputed values where available. The weights were
set to be inversely proportional to the number of observations for each species,
which allowed us to incorporate intraspecic variation while also ensuring that
each species had the same overall contribution to trait relationships. Representative
vectors for each axis were identied by selecting those that loaded most uniquely
on each of the rst two principle component axes.
Abiotic relationships. To explore the relationships between the primary trait axes
and environmental conditions, we used Shapley values from random-forest models.
Shapley values are a game-theoretic metric that partitions the relative inuence of
each variable on the outcome, for every observation in the dataset60,95,96.Itisa
machine-learning analog to partial regression, in that it looks to quantify the
marginal relationship between predictors and outcome variables, taking into
account all other variables in the model. To quantify Shapley values, we used the
random forest to model each of the PC axis values as a function of the ten
environmental variables. We then used the fastshap package in R to estimate the
Shapley values for each variable for each observation using out-of-t test dataset
(15% of data withheld), and we quantied the overall variable importance by taking
the sum of the absolute Shapley values. The environmental variables in Fig. 3a, b
are ranked by variable importance, and the trends for the three variables with the
highest Shapley values are shown in Fig. 3ch for each axis. We also conducted a
second analysis on the full set of 50 covariates to depict the full set of trends across
climate, soil, and topography (Supplementary Figs. 17, 18), but note that many of
these relationships are redundant due to high correlations between variables.
Functional cluster analysis. In order to identify the dimensionality of trait space
and identify the dominant functional trait clusters capturing the full variation in
tree form and function, we conducted hierarchical clustering on the species-level
correlation matrix. First, we calculated species-weighted rank correlations between
pairs of traits using the wCorr package in R, which again allowed us to incorporate
intraspecic trait variation while ensuring each species contributed equal weight.
The optimal number of clusters was identied using the silhouette method in the
dendextend package in R, and the dendrogram was subsequently cut into clusters
based on groups of traits which exhibited consistently high average intra-group
correlation. As an alternate measure of trait correlation which accounts for phy-
logenetic relatedness, we calculated phylogenetic independent contrasts97 (PIC) on
species-level average trait values using the ape package. PIC adjusts for the non-
independence of species due to their shared evolutionary history, and allows us to
remove the effects of pseudo-replication (i.e. among closely related species within a
genus) when calculating correlations. The corresponding correlations among these
contrasts are shown in the bottom triangle of the correlation matrix in Fig. 4a.
Species-level phylogenetic conservatism was calculated using the empirically
measured (non-imputed) trait values via Pagelsλ98,99 which quanties the extent
to which trait correlations among species can be explained by their shared evo-
lutionary history, with a value of one equivalent high phylogenetic signal (under
Brownian motion), and a value of zero equivalent to no phylogenetic signal (a star
NATURE COMMUNICATIONS | (2022) 13:3185 | | 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved
All analyses were conducted in R v. 4.2.0, with the exception of the phylogenetic
eigenvector calculations, which were obtained using the Arpack package in Julia
v. 1.6.2.
Reporting summary. Further information on research design is available in the Nature
Research Reporting Summary linked to this article.
Data availability
The imputed trait data used in this study have been deposited in a dedicated GitHub
repository with (
tree_traits), along with the cleaned phylogenetic eigenvectors from the seed plant
phylogeny of Smith & Brown (2018) and corresponding tree species names matched
using BGCI GlobalTreeSearch database v1.3. The raw TRY trait data are not available
due to data privacy and sharing restrictions, but can be obtained by making a data
request to the TRY Plant Trait Database (
Code availability
The code for replicating the central ndings of this paper is available at https://github.
com/dsmaynard/tree_traits (
Received: 3 November 2021; Accepted: 23 May 2022;
1. Pan, Y., Birdsey, R. A., Phillips, O. L. & Jackson, R. B. The structure,
distribution, and biomass of the worlds forests. Annu. Rev. Ecol. Evol. Syst. 44,
593622 (2013).
2. Pérez-Harguindeguy, N. & Díaz, S. New handbook for standardised
measurement of plant functional traits worldwide. Aust. J. Bot. 61, 167234
3. Kraft, N. J. B., Godoy, O. & Levine, J. M. Plant functional traits and the
multidimensional nature of species coexistence. Proc. Natl Acad. Sci. USA 112,
797802 (2015).
4. Westoby, M. & Wright, I. J. Land-plant ecology on the basis of functional
traits. Trends Ecol. Evol. 21, 261268 (2006).
5. Díaz, S. et al. The global spectrum of plant form and function. Nature 529,
117 (2015).
6. Chave, J. et al. Towards a worldwide wood economics spectrum. Ecol. Lett. 12,
351366 (2009).
7. Maracahipes, L. et al. How to live in contrasting habitats? Acquisitive and
conservative strategies emerge at inter- and intraspecic levels in savanna and
forest woody plants. Perspect. Plant Ecol. Evol. Syst. 34,1725 (2018).
8. Rueda, M., Godoy, O. & Hawkins, B. A. Trait syndromes among North
American trees are evolutionarily conserved and show adaptive value over
broad geographic scales. Ecography 41, 540550 (2018).
9. Kraft, N. J. B. & Ackerly, D. D. Functional trait and phylogenetic tests of
community assembly across spatial scales in an Amazonian forest. Ecol.
Monogr. 80, 401422 (2010).
10. Greenwood, S. et al. Tree mortality across biomes is promoted by drought
intensity, lower wood density and higher specic leaf area. Ecol. Lett. 20,
539553 (2017).
11. Stahl, U., Reu, B. & Wirth, C. Predicting speciesrange limits from functional
traits for the tree ora of North America. Proc. Natl Acad. Sci. USA 111,
1373913744 (2014).
12. Aubin, I. et al. Traits to stay, traits to move: a review of functional traits to
assess sensitivity and adaptive capacity of temperate and boreal trees to
climate change. Environ. Rev. 24, 164186 (2016).
13. Niinemets, Ü. Global-scale climatic controls of leaf dry mass per area, density,
and thickness in trees and shrubs. Ecology 82, 453469 (2001).
14. Wright, I. J. et al. The worldwide leaf economics spectrum. Nature 428,
821827 (2004).
15. Reich, P. B. The world-wide fast-slowplant economics spectrum: a traits
manifesto. J. Ecol. 102, 275301 (2014).
16. Adler, P. B. et al. Functional traits explain variation in plant life history
strategies. Proc. Natl Acad. Sci. USA 111, 10019 (2014).
17. Joswig, J. S. et al. Climatic and soil factors explain the two-dimensional
spectrum of global plant trait variation. Nat. Ecol. Evol. 6,2427 (2022).
18. Poorter, L. et al. The importance of wood traits and hydraulic conductance for
the performance and life history strategies of 42 rainforest tree species. N.
Phytol. 185, 481492 (2010).
19. Wright, S. J. et al. Functional traits and the growth-mortality trade-off in
tropical trees. Ecology 91, 36643674 (2010).
20. Sakschewski, B. et al. Leaf and stem economics spectra drive diversity of
functional plant traits in a dynamic global vegetation model. Glob. Chang.
Biol. 21, 27112725 (2015).
21. Petit, R. J. & Hampe, A. Some evolutionary consequences of being a tree.
Annu. Rev. Ecol. Evol. Syst. 37, 187214 (2006).
22. Swenson, N. G. & Weiser, M. D. Plant geography upon the basis of functional
traits: an example from eastern North American trees. Ecology 91, 22342241
23. Fan, Y., Miguez-Macho, G., Jobbágy, E. G., Jackson, R. B. & Otero-Casal, C.
Hydrologic regulation of plant rooting depth. Proc. Natl Acad. Sci. USA 114,
1057210577 (2017).
24. Purves, D. W., Lichstein, J. W. & Pacala, S. W. Crown plasticity and
competition for canopy space: a new spatially implicit model parameterized
for 250 North American tree species. PLoS ONE 2, e870 (2007).
25. Harja, D., Vincent, G., Mulia, R. & van Noordwijk, M. Tree shape plasticity in
relation to crown exposure. Trees Struct. Funct. 26, 12751285 (2012).
26. Moles, A. T. et al. Factors that shape seed mass evolution. Proc. Natl Acad. Sci.
USA 102, 1054010544 (2005).
27. Wieczynski, D. J. et al. Climate shapes and shifts functional biodiversity in
forests worldwide. Proc. Natl Acad. Sci. USA 116, 587592 (2019).
28. Pinho, B. X. et al. Functional biogeography of Neotropical moist forests:
traitclimate relationships and assembly patterns of tree communities. Glob.
Ecol. Biogeogr. 30, 14301446 (2021).
29. Kattge, J. et al. TRY plant trait databaseenhanced coverage and open access.
Glob. Chang. Biol. 26, 119188 (2020).
30. Schrodt, F. et al. BHPMFa hierarchical Bayesian approach to gap-lling and
trait prediction for macroecology and functional biogeography. Glob. Ecol.
Biogeogr. 24, 15101521 (2015).
31. Penone, C. et al. Imputation of missing data in life-history trait datasets: which
approach performs the best? Methods Ecol. Evol. 5,110 (2014).
32. Bauman, D., Drouet, T., Dray, S. & Vleminckx, J. Disentangling good from
bad practices in the selection of spatial or phylogenetic eigenvectors.
Ecography 41, 16381649 (2018).
33. Gatti, R. C. et al. The number of tree species on Earth. Proc. Natl Acad. Sci.
USA 119,111 (2022).
34. Mill, R. R. et al. A new classication and linear sequence of extant
gymnosperms. Phytotaxa 70,5570 (2011).
35. BGCI. GlobalTreeSearch online database (version 1.4).Botanic Gardens
Conservation International.
36. Zanne, A. E. et al. Functional biogeography of angiosperms: life at the
extremes. N. Phytol. 218, 16971709 (2018).
37. Karger, D. N. et al. Climatologies at high resolution for the earths land surface
areas. Sci. Data 4,120 (2017).
38. Karger, D. N., Conrad, O., Böhner, J., Kawohl, T. & Kreft, H. Data from:
climatologies at high resolution for the earths land surface areas. Dryad Digit.
Repos. (2018).
39. Trabucco, A. & Zomer, R. Global Aridity Index and Potential
Evapotranspiration (ET0) Climate Database v2. Figshare.
6084/m9.gshare.7504448.v3 (2019).
40. Obu, J. et al. Northern Hemisphere permafrost map based on TTOP
modelling for 20002016 at 1/km2 scale. Earth Sci. Rev. 193, 299316 (2019).
41. Hengl, T. et al. SoilGrids250m: global gridded soil information based on
machine learning. PLoS ONE 12,140 (2017).
42. Tuanmu, M. N. & Jetz, W. A global, remote sensing-based characterization of
terrestrial habitat heterogeneity for biodiversity and ecosystem modelling.
Glob. Ecol. Biogeogr. 24, 13291339 (2015).
43. Fan, Y., Li, H. & Miguez-Macho, G. Global patterns of groundwater table
depth. Science 339, 940943 (2013).
44. Diniz-Filho, J. A. F., De SanTAna, C. E. R. & Bini, L. M. An eigenvector
method for estimating phylogenetic inertia. Evolution 52, 12471262 (1998).
45. Swenson, N. G. Phylogenetic imputation of plant functional trait databases.
Ecography 37, 105110 (2014).
46. Roberts, D. R. et al. Cross-validation strategies for data with temporal, spatial,
hierarchical, or phylogenetic structure. Ecography 40, 913929 (2017).
47. Wadoux, A. M. J., Heuvelink, G. B. M., Bruin, S., De & Brus, D. J. Spatial
cross-validation is not the right way to evaluate map accuracy. Ecol. Modell.
457, 109692 (2021).
48. Meinshausen, N. Quantile Regression forests. J. Mach. Learn. Res. 7, 983999
49. Poorter, H., Niinemets, Ü., Poorter, L., Wright, I. J. & Villar, R. Causes and
consequences of variation in leaf mass per area (LMA): a meta-analysis. N.
Phytol. 182, 565588 (2009).
50. Onoda, Y. et al. Global patterns of leaf mechanical properties. Ecol. Lett. 14,
301312 (2011).
10 NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
51. Prach, K. & Walker, L. R. Four opportunities for studies of ecological
succession. Trends Ecol. Evol. 26, 119123 (2011).
52. He, Z. et al. Metagenomic analysis reveals a marked divergence in the
structure of belowground microbial communities at elevated CO2. Ecol. Lett.
13, 564575 (2010).
53. Witkowski, E. T. F. & Lamont, B. B. Leaf specic mass confounds leaf density
and thickness. Oecologia 88, 486493 (1991).
54. Niinemets, Ü. A review of light interception in plant stands from leaf to
canopy in different plant functional types and in species with varying shade
tolerance. Ecol. Res. 25, 693714 (2010).
55. Givnish, T. J. Plant stems: biomechanical adaptation for energy capture and
inuence on species distributions. In Plant Stems: Physiology and Functional
Morphology (ed. Gartner, B. L.) 349 (Academic Press, 1995).
56. King, D. A., Davies, S. J., Tan, S. & Noor, N. S. M. The role of wood density
and stem support costs in the growth and mortality of tropical trees. J. Ecol.
94, 670680 (2006).
57. Simard, M., Pinto, N., Fisher, J. B. & Baccini, A. Mapping forest canopy height
globally with spaceborne lidar. J. Geophys. Res. Biogeosci. 116,112 (2011).
58. Klein, T., Randin, C. & Körner, C. Water availability predicts forest canopy
height at the global scale. Ecol. Lett. 18, 13111320 (2015).
59. Choat, B. et al. Global convergence in the vulnerability of forests to drought.
Nature 491, 752755 (2012).
60. Štrumbelj, E. & Kononenko, I. Explaining prediction models and individual
predictions with feature contributions. Knowl. Inf. Syst. 41, 647665 (2014).
61. Moles, A. T. et al. Which is a better predictor of plant traits: temperature or
precipitation? J. Veg. Sci. 25, 11671180 (2014).
62. Mittelbach, G. G. et al. Evolution and the latitudinal diversity gradient:
speciation, extinction and biogeography. Ecol. Lett. 10, 315331 (2007).
63. Schemske, D. W., Mittelbach, G. G., Cornell, H. V., Sobel, J. M. & Roy, K. Is
there a latitudinal gradient in the importance of biotic interactions? Annu.
Rev. Ecol. Evol. Syst. 40, 245269 (2009).
64. Maire, V. et al. Global effects of soil and climate on leaf photosynthetic traits
and rates. Glob. Ecol. Biogeogr. 24, 706717 (2015).
65. Paulsen, J., Weber, U. M. & Körner, C. Tree growth near treeline: abrupt or
gradual reduction with altitude? Arct. Antarct. Alp. Res. 32,1420 (2000).
66. Fajardo, A. Insights into intraspecic wood density variation and its
relationship to growth, height and elevation in a treeline species. Plant Biol.
20, 456464 (2018).
67. Tripler, C. E., Kaushal, S. S., Likens, G. E. & Todd Walter, M. Patterns in
potassium dynamics in forest ecosystems. Ecol. Lett. 9, 451466 (2006).
68. Wright, I. J. et al. Global climatic drivers of leaf size. Science 357, 917921
69. Williams, V. L., Witkowski, E. T. F. & Balkwill, K. Relationship between bark
thickness and diameter at breast height for six tree species used medicinally in
South Africa. South Afr. J. Bot. 73, 449465 (2007).
70. Rosell, J. A. Bark thickness across the angiosperms: more than just re. N.
Phytol. 211,90102 (2016).
71. Schafer, J. L., Breslow, B. P., Hohmann, M. G. & Hoffmann, W. A. Relative
bark thickness is correlated with tree species distributions along a re
frequency gradient. Fire Ecol. 11,7487 (2015).
72. Rosell, J. A. et al. Inner bark as a crucial tissue for non-structural carbohydrate
storage across three tropical woody plant communities. Plant Cell Environ. 44,
156170 (2021).
73. Hodgson, J. G. et al. Is leaf dry matter content a better predictor of soil fertility
than specic leaf area? Ann. Bot. 108, 13371345 (2011).
74. Weemstra, M. et al. Towards a multidimensional root trait framework: a tree
root review. N. Phytol. 211, 11591169 (2016).
75. Laughlin, D. C. et al. Root traits explain plant species distributions along
climatic gradients yet challenge the nature of ecological trade-offs. Nat. Ecol.
Evol. 5, 11231134 (2021).
76. Chen, G., Yang, Y. & Robinson, D. Allometric constraints on, and trade-offs
in, belowground carbon allocation and their control of soil respiration across
global forest ecosystems. Glob. Chang. Biol. 20, 16741684 (2014).
77. Wilfahrt, P. A., Collins, B. & White, P. S. Shifts in functional traits among tree
communities across succession in eastern deciduous forests. Forest Ecol.
Manag. 324, 179185 (2014).
78. Krishnadas, M. et al. Environment and past land use together predict
functional diversity in a temperate forest. Ecol. Appl. 28, 21422152 (2018).
79. Bradford, M. A. et al. Climate fails to predict wood decomposition at regional
scales. Nat. Clim. Chang. 4, 625630 (2014).
80. R Core Team. R: A Language and Environment for Statistical Computing.(R
Foundation for Statistical Computing, 2021).
81. Smith, S. A. & Brown, J. W. Constructing a broadly inclusive seed plant
phylogeny. Am. J. Bot. 105, 302314 (2018).
82. Beech, E., Rivers, M., Oldeld, S. & Smith, P. P. GlobalTreeSearch: the rst
complete global database of tree species and country distributions. J. Sustain.
Forest. 36, 454489 (2017).
83. Funk, J. L. The physiology of invasive plants in low-resource environments.
Conserv. Physiol. 1,117 (2013).
84. Sperry, J. S., Hacke, U. G. & Pittermann, J. Size and function in conifer
tracheids and angiosperm vessels. Am. J. Bot. 93, 14901500 (2006).
85. Wright, M. N. & Ziegler, A. Ranger: a fast implementation of random forests
for high dimensional data in C++ and R. J. Stat. Softw. 77,117 (2017).
86. Breiman, L. Random forests. Mach. Learn. 45,532 (2001).
87. Probst, P., Wright, M. N. & Boulesteix, A. L. Hyperparameters and tuning
strategies for random forest. Wiley Interdiscip. Rev. Data Min. Knowl. Discov.
9,119 (2019).
88. Gschwantner, T. et al. Common tree denitions for national forest inventories
in Europe. Silva Fenn. 43, 303321 (2009).
89. Fox, E. W., Ver Hoef, J. M. & Olsen, A. R. Comparing spatial regression to
random forests for large environmental data sets. PLoS ONE 15,122 (2020).
90. van den Hoogen, J. et al. Soil nematode abundance and functional group
composition at a global scale. Nature 572, 194198 (2019).
91. Diniz-Filho, J. A. F., Rangel, T. F., Santos, T. & Mauricio Bini, L. Exploring
patterns of interspecic variation in quantitative traits using sequential
phylogenetic eigenvector regressions. Evolution 66, 10791090 (2012).
92. Stekhoven, D. J. & Bühlmann, P. Missforest-Non-parametric missing value
imputation for mixed-type data. Bioinformatics 28, 112118 (2012).
93. Dormann, C. F. et al. Methods to account for spatial autocorrelation in the
analysis of species distributional data: a review. Ecography 30, 609628 (2007).
94. Li, J. Assessing the accuracy of predictive models for numerical data: Not r nor
r2, why not? Then what? PLoS ONE 12,116 (2017).
95. Shapley, L. S. A value for n-person games. Contributions to the Theory of
Games II 2.28 (eds Kuhn, H. W. & Tucker, A. W.) 307317 (Princeton
University Press, 1953).
96. Marcilio, W. E. & Eler, D. M. From explanations to feature selection: assessing
SHAP values as feature selection mechanism. Proc. - 2020 33rd SIBGRAPI Conf.
Graph. Patterns Images 340347,
00053 (2020).
97. Felsenstein, J. Phylogenies and the comparative method. Am. Nat. 125,115
98. Pagel, M. Inferring the historical patterns of biological evolution. Nature 401,
877884 (1999).
99. Münkemüller, T. et al. How to measure and test phylogenetic signal. Methods
Ecol. Evol. 3, 743756 (2012).
This work was primarily funded by the Swiss National Science Foundation, Ambizione
grant #PZ00P3_193612 to D.S.M. Additional support was provided by D.O.B. Ecology to
T.W.C.; SNSF Ambizione grant #PZ00P3_193646 to C.M.Z.; SNSF Ambizione grant #
PZ00P3_179900 to C.A.; U.S. National Science Foundation Graduate Research Fellow-
ship and Embassy of Switzerland in the USA ThinkSwiss Research Fellowship to G.R.S.
and UNAM-PAPIIT grant #IN210220 to J.A.R. We thank Ángel de Frutos, Gabriela
Lopez-Gonzalez, Owen Atkin and Yongfu Chai for sharing data, as well as the TRY
database managers and numerous researchers who contributed open-access data.
Author contributions
D.S.M. conceived of this study, analyzed the data and wrote the rst draft of the
manuscript, with assistance from L.B., C.M.Z., C.A., J.v.d.H., H.M., L.M., G.R.S. and
T.W.C. Data were contributed by I.A., E.B., C.C.F.B., J.A.C., B.E.L.C., A.T.R.A., B.X.P.,
A.S.D., A.G.-M., P.H., C.H.L., A.S.K., Ü.N., V.D.P., J.A.R., F.M.S., S.S., A.C.d.S., Ê.S.,
P.M.v.B., E.W., G.B. and J.K. All authors provided suggestions and feedback on the
analyses and interpretations, and contributed to the writing and revising of the
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material
available at
Correspondence and requests for materials should be addressed to Daniel S. Maynard.
Peer review information Nature Communications thanks the anonymous reviewer(s) for
their contribution to the peer review of this work.
Reprints and permission information is available at
Publishers note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional afliations.
NATURE COMMUNICATIONS | (2022) 13:3185 | | 11
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the articles Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
articles Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit
© The Author(s) 2022
Institute of Integrative Biology, ETH Zürich, 8092 Zürich, Switzerland.
Department of Biology, Stanford University, Stanford, CA 94305, USA.
Department of Science, Roma Tre University, Rome, Italy.
Natural Resources Canada, Canadian Forest Service, Great Lakes Forestry Centre,
Sault Ste Marie, ON P6A 2E5, Canada.
Environmental Change Institute, School of Geography and the Environment, University of Oxford,
Oxford, UK.
Lancaster Environment Centre, Lancaster University, Lancaster, UK.
Department of Aquatic Ecology & Environmental Biology,
Institute for Water and Wetland Research, Radboud University, Nijmegen, The Netherlands.
Department of Geography, Kings College London,
30 Aldwych, London WC2B 4BG, UK.
Department of Biotechnologies and Life Sciences (DBSV), University of Insubria, 21100 Varese, Italy.
Goethe University, Institute for Physical Geography, Altenhöferallee 1, 60438 Frankfurt am Main, Germany.
Biology Department, Faculty of
Natural Sciences, Universidad del Rosario, Avenida carrera 24 # 63C-69, Bogotá, Colombia.
Institute of Botany, University of Natural Resources
and Life Sciences, Gregor Mendel St. 33, 1190 Vienna, Austria.
Environmental Research Institute, University of Waikato, Hamilton, New Zealand.
Research Center for Advanced Science and Technology, The University of Tokyo, 4-6-1 Komaba, Meguro, Tokyo 153-8904, Japan.
Chair of
Crop Science and Plant Biology, Estonian University of Life Sciences, Tartu 51006, Estonia.
Department of Ecology, Universidade Federal do Rio
Grande do Sul, Porto Alegre, RS 91501-970, Brazil.
AMAP, Univ Montpellier, INRAe, CIRAD, CNRS, IRD, Montpellier, France.
Departamento de
Botânica, Universidade Federal de Pernambuco, Recife, Pernambuco, Brazil.
Laboratorio Nacional de Ciencias de la Sostenibilidad, Instituto de
Ecología, Universidad Nacional Autónoma de México, A.P. 70-275, Ciudad Universitaria, Coyoacán, 04510 Mexico City, Mexico.
Institute of
Landscape and Plant Ecology, University of Hohenheim, Ottilie-Zeller-Weg 2, D-70599 Stuttgart, Germany.
Komarov Botanical Institute, Prof.
Popov str., 2, St. Petersburg 197376, Russia.
Department of Forestry, Santa Catarina State University, Lages, SC 88520-000, Brazil.
Clima Temperado, Pelotas, RS 96010-971, Brazil.
Institute of Environmental Science, Leiden University, 2333 CC Leiden, the Netherlands.
Department of Biology, University of Wisconsin Eau Claire, Eau Claire, WI 54702, USA.
Max Planck Institute for Biogeochemistry, 07745
Jena, Germany.
German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, 04103 Leipzig, Germany.
12 NATURE COMMUNICATIONS | (2022) 13:3185 | |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
... In addition, we created distance measures by calculating the spherical distance to shipping ports 81 and airports 82 . All layers were standardized to a 30 arcsec resolution (~1 km 2 at the equator), a resolution at which these variables have been shown to have an influence on plant biogeography and assembly patterns 83,84 . We chose model variables to represent both climate and soil properties that exhibited low collinearity for each of three datasets: global (all 14 biomes from Supplementary Table 1), temperate (temperate broadleaf, coniferous, grassland biomes) and tropical (tropical moist broadleaf, deciduous broadleaf, coniferous, and grassland biomes). ...
... For functional diversity, we calculated the analogous metrics using trait distance matrices instead of phylogenetic tree-based distances. We selected eight traits extracted from Maynard et al. 83 that represented the major clusters of functional trait diversity, thereby capturing the full spectrum of tree form and function while minimizing correlation between traits. Maynard et al. 83 used data from the TRY plant trait database to parametrize machine learning models to estimate the expression of 18 traits as a function of the local environment and/or phylogeny. ...
... We selected eight traits extracted from Maynard et al. 83 that represented the major clusters of functional trait diversity, thereby capturing the full spectrum of tree form and function while minimizing correlation between traits. Maynard et al. 83 used data from the TRY plant trait database to parametrize machine learning models to estimate the expression of 18 traits as a function of the local environment and/or phylogeny. The observed trait data underlying these models encompassed 491,001 unique observations across 13,189 species from 2,313 genera, with consistent representation across taxonomic orders. ...
Full-text available
In Xavier Bosch's interview “Spain's science minister sees future in telecom” (News Focus, 31 Jan., p. 653), Josep Pique, Spanish Minister of Science and Technology, asserts that “[n]ow there are many more scientists from abroad working in Spain than there are Spanish scientists abroad.”
... Conservative species included coniferous species that are longer-lived, favour slower growth and resource acquisition (Maynard et al., 2022) and often display more conservative hydraulic strategies than acquisitive species (Reich, 2014;Yang et al., 2022). Coniferous species have lower stem conductivity and slow-decomposing litter with smaller, thicker leaves (Díaz et al., 2016). ...
... . Interactions between disturbance factors are complex, and often result in disparate effects on tree demography in temperate forests(Benito-Garzón et al., 2013;Millar & Stephenson, 2015;van Mantgem et al., 2013;van Mantgem & Stephenson, 2007), as observed in this study.Results also highlighted a variation in the direction of species responses to the interactive effects that is likely the result of divergent life history strategies. Species included in this study reflect a wellknown fast-slow plant economic spectrum(Reich, 2014) describing a trade-off between acquisition and conservation of resources that ultimately determine species ecological strategies(Maynard et al., 2022). Under water stress, plants with greater water acquisition and conservation should be selected against plants that acquire and use water rapidly and often experience high mortality in drier conditions(Montwé et al., 2015). ...
Full-text available
Studies of drivers shaping forest communities frequently include abiotic or biotic factors, while their interactive effects remain understudied. Here, we combined data on two prominent abiotic gradients, climatic moisture deficit and wildfire probability, along with tree functional neighbourhoods (i.e. trait differences of close neighbours) to assess variation in survival and growth of 56 tree species in the northwestern US. We asked two questions: (1) How does functional neighbourhood dissimilarity vary with environmental gradients? and (2) How do demographic rates of tree species in the northwestern United States vary with the interactive effects of environmental gradients and functional neighbourhood? We expected functional neighbourhoods to become more similar as environmental stress increased, due to a convergence of species towards an optimum stress tolerance strategy. We also predicted the interactive effects of abiotic and biotic factors on tree demography and high variation in species‐specific responses to these interactive effects due to divergent species life history strategies. Functional neighbourhoods defined by dissimilarities in stem conductivity, litter decomposition, resprouting ability and specific leaf area changed with climate, shifting to more diverse neighbourhoods as climatic moisture deficit and wildfire probability increased. Results supported interactive effects of the functional neighbourhood and climatic moisture deficit or wildfire probability on tree demography, but only when the identity of dominant species was considered. Species‐specific responses were highly variable in their direction and magnitude and often demonstrated opposing effects of climate and the functional neighbourhood and climatic moisture deficit on tree demography. Synthesis . Our findings show that climate and tree neighbourhood functional dissimilarity jointly impact tree demography; however, the effects are species‐specific. Results of this study highlight the need to consider the interactive effects of abiotic and biotic contexts and individual species responses to their environment to adequately understand tree persistence under current and future climate conditions.
... Further, plant functional traits show different responses to a variable environment and neighbors, increasing efficient utilization of resources (Laughlin et al., 2020;Stotz et al., 2021). Plant height, stem diameter, and leaf economic traits related to light-use efficiency, affect interspecific interactions and change community assembly (Roscher et al., 2015;Williams et al., 2020;Maynard et al., 2022). Root functional traits are sensitive to soil resource availability and increase the acquisition of water or nutrients through plasticity of specific root length, root length density and root tissue density (Wurzburger and Wright, 2015;Yu et al., 2022). ...
... Single functional traits show correlations with average fitness differences between species (Kraft et al., 2015). The correlation between functional traits shows different strategies of species in forests (Maynard et al., 2022), but the complex relationship (synergies or trade-offs) between different functional traits remains poorly understood in agroecosystems. ...
... In recent years, considerable research has been conducted on the stoichiometric characteristics of plant leaves, and the relationships between leaf stoichiometric characteristics and biological and abiotic factors have been explored at the regional [10,11], national [12,13], and global scales [14,15]. Despite extensive research on the relationship between traits in the plant kingdom, our understanding of the coordination of traits at the level of woody organisms lags behind that of herbaceous species [16]. Previous studies have identified a set of key functional traits that summarize the form and functional spectrum of the entire plant kingdom, with leaf economy and plant size being the main trait axes supporting life cycle strategies [16]. ...
... Despite extensive research on the relationship between traits in the plant kingdom, our understanding of the coordination of traits at the level of woody organisms lags behind that of herbaceous species [16]. Previous studies have identified a set of key functional traits that summarize the form and functional spectrum of the entire plant kingdom, with leaf economy and plant size being the main trait axes supporting life cycle strategies [16]. Due to the size, life span, individual development and unique structural characteristics of woody plants, compared with herbaceous plants, trees have obvious characteristics and face new abiotic stress [17]. ...
Full-text available
The main research content of the field of ecological stoichiometry is the energy of various chemical elements and the interaction between organisms and the environment throughout ecological processes. Nitrogen and phosphorus are the main elements required for the growth and development of plants and these also form the constituent basis of biological organisms. Both elements interact and jointly regulate the growth and development of plants, and their element ratios are an indication of the nutrient utilization rate and nutrient limitation status of plants. Previous research developed a general biogeography model of the stoichiometric relationship between nitrogen and phosphorus in plant leaves on a global scale. Further, it was shown that the relative rate of nitrogen uptake by leaves was lower than that of phosphorus, and the scaling exponent of nitrogen and phosphorus was 2/3. However, it is not clear how the stoichiometric values of nitrogen and phosphorus, especially their scaling exponents, change in the leaves of Chinese woody plants in response to changing environmental conditions. Therefore, data sets of leaf nitrogen and phosphorus concentrations, and nitrogen to phosphorus ratios in Chinese woody plants were compiled and classified according to different life forms. The overall average concentrations of nitrogen and phosphorus in leaves were 20.77 ± 8.12 mg g ⁻¹ and 1.58 ± 1.00 mg g ⁻¹ , respectively. The contents of nitrogen and phosphorus in leaves of deciduous plants were significantly higher than those of evergreen plants. In leaves, life form is the main driving factor of nitrogen content, and mean annual temperature is the main driving factor of phosphorus content; soil available nitrogen is the main driving factor of the nitrogen to phosphorus ratio. These values can be used for comparison with other studies. In addition, the scale index was found to be significantly different among different life forms. The scaling exponents of N-P of woody plants of different life forms, such as trees, shrubs, evergreen, deciduous, and coniferous plants are 0.67, 0.72, 0.63, 0.72, and 0.66, respectively. The N-P scaling exponent of shrubs was higher than that of trees, and that of deciduous plants was higher than that of evergreen plants. These results suggest that the internal attributes of different life forms, the growth rate related to phosphorus, and the relative nutrient availability of soil are the reasons for the unsteady relationship between nitrogen and phosphorus in leaves.
... Plant traits associated with the whole-plant economic spectrum suggest that essential patterns of form and function across species can be captured by the two-dimensional spectra of multiple trait combinations (Lavorel and Garnier, 2002;McGill et al., 2006;Westoby and Wright, 2006;Diaz et al., 2015;Laughlin and Messier, 2015;Joswig et al., 2022). As such, quantifying plant trait spectrum space can better capture the breadth and adaptability of tree form and function (Maynard et al., 2022). This old-growth forest across species demonstrates that multiple trait combinations are fundamentally shaped by two main trade-off axes: resource acquisition-conservation and hydraulic safety-efficiency ( Figure 3A). ...
Full-text available
Climate change leads to novel species interactions and continues to reshuffle ecological communities, which significantly declines carbon accumulation rates in mature forests. Still, little is known about the potential influence of multiple global change factors on long-term biomass dynamics and functional trait combinations. We used temporal demographic records spanning 26 years and extensive databases of functional traits to assess how old-growth subtropical forest biomass dynamics respond to various climatic change scenarios (extreme drought, subsequent drought, warming, elevated CO 2 concentrations, and windstorm). We found that the initial severe drought, subsequent drought and windstorm events increased biomass loss due to tree mortality, which exceeded the biomass gain produced by survivors and recruits, ultimately resulting in more negative net biomass balances. These drought and windstorm events caused massive biomass loss due to tree mortality that tended towards acquisition species with high hydraulic efficiency, whereas biomass growth from survivors and recruits tended to consist of acquisition species with high hydraulic safety. Compensatory growth in this natural forest provided good explanation for the increase in biomass growth after drought and windstorm events. Notably, these dominant-species transitions reduced carbon storage and residence time, forming a positive carbon-climate feedback loop. Our findings suggest that climate changes could alter functional strategies and cause shifts in new dominant species, which could greatly reduce ecological functions and carbon gains of old-growth subtropical forests.
Full-text available
Climate change is rapidly altering weather patterns, resulting in shifts in climatic zones. The survival of trees in specific locations depends on their functional traits. Local populations exhibit trait adaptations that ensure their survival and accomplishment of growth and reproduction processes during the growing season. Studying these traits offers valuable insights into species responses to present and future environmental conditions, aiding the implementation of measures to ensure forest resilience and productivity. This study investigates the variability in functional traits among five black spruce ( Picea mariana (Mill.) B.S.P.) provenances originating from a latitudinal gradient along the boreal forest, and planted in a common garden in Quebec, Canada. We examined differences in bud phenology, growth performance, lifetime first reproduction, and the impact of a late-frost event on tree growth and phenological adjustments. The findings revealed that trees from northern sites exhibit earlier budbreak, lower growth increments, and reach reproductive maturity earlier than those from southern sites. Late-frost damage affected growth performance, but no phenological adjustment was observed in the successive year. Local adaptation in the functional traits may lead to maladaptation of black spruce under future climate conditions or serve as a potent evolutionary force promoting rapid adaptation under changing environmental conditions.
Full-text available
Background and aims Coordination between leaf and root traits is crucial to plant performance and ecosystem functioning, but how leaves and roots coordinate in ectomycorrhizal (ECM)-dominated alpine forests remains unclear. Therefore, the covariation patterns of leaf and root traits of ECM-dominated alpine conifers and the environmental drivers were examined. Methods Five pairs of key leaf (i.e., leaf thickness [LT], specific leaf area [SLA], leaf tissue density [LTD], leaf N and P concentrations) and fine-root traits (i.e., root diameter [RD], specific root length [SRL], root tissue density [RTD], root N and P concentrations) were measured across 49 alpine coniferous populations (including 8 coniferous species) on the Tibetan Plateau. Results Root traits including RTD, root N and P concentrations and leaf traits such as LT, SLA, LTD, leaf N and P concentrations were correlated. The root-leaf relationships represent a tradeoff between resource conservation and fast plant growth, i.e., plant economic spectrum. RD and SRL were independent from the plant economic spectrum. Temperature drove variations in the leaf traits, RTD, root N and P concentrations, and conifers under low temperature had denser leaves and roots (i.e., larger LT, LTD, RTD) and lower nutrient contents. Precipitation primarily controlled variations in RD and SRL, and roots became thinner with decreasing precipitation. Conclusion Our study demonstrates divergent roles of temperature and precipitation in driving the coordination of leaf and root economic traits in the ECM-dominated alpine coniferous ecosystems. This is insightful for a comprehensive understanding of the adaptation and responses of alpine forests to climate change.
Full-text available
The metabolome is the biochemical basis of plant form and function, but we know little about its macroecological variation across the plant kingdom. Here, we used the plant functional trait concept to interpret leaf metabolome variation among 457 tropical and 339 temperate plant species. Distilling metabolite chemistry into five metabolic functional traits reveals that plants vary on two major axes of leaf metabolic specialization-a leaf chemical defense spectrum and an expression of leaf longevity. Axes are similar for tropical and temperate species, with many trait combinations being viable. However, metabolic traits vary orthogonally to life-history strategies described by widely used functional traits. The metabolome thus expands the functional trait concept by providing additional axes of metabolic specialization for examining plant form and function.
Tree plantations can facilitate the establishment of native trees, thus providing opportunities for the ecological restoration or rehabilitation of degraded forest lands. Stand variables can influence the establishment of native tree species, but few studies have been carried out to determine how sapling growth response to stand variables varies among functionally different species. We evaluated the effect of stand age, stand basal area, tree density, and time since last logging on stem diameter growth rates, as well as the dependence of such effect on whole plant, stem, and leaf functional traits. We measured the stem diameter increment on 280 individuals of 22 species in permanent plots for a period of 2 years in araucaria ( Araucaria angustifolia (Bertol.) Kuntze) plantations in the Atlantic Forest, NE Argentina. Increasing all-species basal area and plantation age negatively affected the growth rates, and such effect tended to be higher in species with low wood density. Two functional axes independently explained growth rate differences between species in growth rate. Growth rates increase with decreasing wood density, increasing leaf phosphorus and potassium content, and decreasing specific leaf area. Among species that are similar in the traits associated with the first functional axis, plant growth increases with increasing leaf nitrogen content.
Full-text available
One of the most fundamental questions in ecology is how many species inhabit the Earth. However, due to massive logistical and financial challenges and taxonomic difficulties connected to the species concept definition, the global numbers of species, including those of important and well-studied life forms such as trees, still remain largely unknown. Here, based on global ground-sourced data, we estimate the total tree species richness at global, continental, and biome levels. Our results indicate that there are ∼73,000 tree species globally, among which ∼9,000 tree species are yet to be discovered. Roughly 40% of undiscovered tree species are in South America. Moreover, almost one-third of all tree species to be discovered may be rare, with very low populations and limited spatial distribution (likely in remote tropical lowlands and mountains). These findings highlight the vulnerability of global forest biodiversity to anthropogenic changes in land use and climate, which disproportionately threaten rare species and thus, global tree richness.
Full-text available
Significance ΔNp63 is a master regulator of skin homeostasis since it finely controls keratinocyte differentiation and proliferation. Here, we provide cellular and molecular evidence demonstrating the functional role of a ΔNp63 interactor, the R-loop–resolving enzyme Senataxin (SETX), in fine-tuning keratinocyte differentiation. We found that SETX physically binds the p63 DNA–binding motif present in two early epidermal differentiation genes, Keratin 1 (KRT1) and ZNF750, facilitating R-loop removal over their 3′ ends and thus allowing efficient transcriptional termination and gene expression. These molecular events translate into the inability of SETX-depleted keratinocytes to undergo the correct epidermal differentiation program. Remarkably, SETX is dysregulated in cutaneous squamous cell carcinoma, suggesting its potential involvement in the pathogenesis of skin disorders.
Full-text available
Significance Tree diversity is fundamental for forest ecosystem stability and services. However, because of limited available data, estimates of tree diversity at large geographic domains still rely heavily on published lists of species descriptions that are geographically uneven in coverage. These limitations have precluded efforts to generate a global perspective. Here, based on a ground-sourced global database, we estimate the number of tree species at biome, continental, and global scales. We estimated a global tree richness (≈73,300) that is ≈14% higher than numbers known today, with most undiscovered species being rare, continentally endemic, and tropical or subtropical. These results highlight the vulnerability of global tree species diversity to anthropogenic changes.
Full-text available
Plant functional traits can predict community assembly and ecosystem functioning and are thus widely used in global models of vegetation dynamics and land–climate feedbacks. Still, we lack a global understanding of how land and climate affect plant traits. A previous global analysis of six traits observed two main axes of variation: (1) size variation at the organ and plant level and (2) leaf economics balancing leaf persistence against plant growth potential. The orthogonality of these two axes suggests they are differently influenced by environmental drivers. We find that these axes persist in a global dataset of 17 traits across more than 20,000 species. We find a dominant joint effect of climate and soil on trait variation. Additional independent climate effects are also observed across most traits, whereas independent soil effects are almost exclusively observed for economics traits. Variation in size traits correlates well with a latitudinal gradient related to water or energy limitation. In contrast, variation in economics traits is better explained by interactions of climate with soil fertility. These findings have the potential to improve our understanding of biodiversity patterns and our predictions of climate change impacts on biogeochemical cycles.
Full-text available
For decades scientists have produced maps of biological, ecological and environmental variables. These studies commonly evaluate the map accuracy through cross-validation with the data used for calibrating the underlying mapping model. Recent studies, however, have argued that cross-validation statistics of most mapping studies are optimistically biased. They attribute these overoptimistic results to a supposed serious methodological flaw in standard cross-validation methods, namely that these methods ignore spatial autocorrelation in the data. They argue that spatial cross-validation should be used instead, and contend that standard cross-validation methods are inherently invalid in a geospatial context because of the autocorrelation present in most spatial data. Here we argue that these studies propagate a widespread misconception of statistical validation of maps. We explain that unbiased estimates of map accuracy indices can be obtained by probability sampling and design-based inference and illustrate this with a numerical experiment on large-scale above-ground biomass mapping. In our experiment, standard cross-validation (i.e., ignoring autocorrelation) led to smaller bias than spatial cross-validation. Standard cross-validation was deficient in case of a strongly clustered dataset that had large differences in sampling density, but less so than spatial cross-validation. We conclude that spatial cross-validation methods have no theoretical underpinning and should not be used for assessing map accuracy, while standard cross-validation is deficient in case of clustered data. Model-free, design-unbiased and valid accuracy assessment is achieved with probability sampling and design-based inference. It is valid without the need to explicitly incorporate or adjust for spatial autocorrelation and perfectly suited for the validation of large scale biological, ecological and environmental maps.
Full-text available
Ecological theory is built on trade-offs, where trait differences among species evolved as adaptations to different environments. Trade-offs are often assumed to be bidirectional, where opposite ends of a gradient in trait values confer advantages in different environments. However, unidirectional benefits could be widespread if extreme trait values confer advantages at one end of an environmental gradient, whereas a wide range of trait values are equally beneficial at the other end. Here, we show that root traits explain species occurrences along broad gradients of temperature and water availability, but model predictions only resembled trade-offs in two out of 24 models. Forest species with low specific root length and high root tissue density (RTD) were more likely to occur in warm climates but species with high specific root length and low RTD were more likely to occur in cold climates. Unidirectional benefits were more prevalent than trade-offs: for example, species with large-diameter roots and high RTD were more commonly associated with dry climates, but species with the opposite trait values were not associated with wet climates. Directional selection for traits consistently occurred in cold or dry climates, whereas a diversity of root trait values were equally viable in warm or wet climates. Explicit integration of unidirectional benefits into ecological theory is needed to advance our understanding of the consequences of trait variation on species responses to environmental change. Scrutinizing the empirical evidence for bidirectional trade-offs in fine root traits, the authors show that while these are important in explaining species occurrences along broad temperature and water availability gradients, unidirectional benefits are prevalent.
Natural climatic treelines are relatively discrete boundaries in the landscape established at a certain elevation within an otherwise continuous gradient of environmental change. By studying tree rings along elevational transects at and below the upper treeline in the European Alps, we (1) determine whether radial stem growth declines abruptly or gradually, and (2) test climatic influences on trees near treeline by investigating transects for climatically different historical periods. While tree height decreases gradually toward the treeline, there is no such general trend for radial tree growth. We found rather abrupt changes which imply threshold effects of temperature which moved upslope in a wave-like manner as temperatures increased over the past 150 yr. Currently radial tree growth at treeline in the Alps is the same magnitude as at several hundred meters below current treeline. Over short intervals, tree-ring width is more dependent on interannual climatic variability than on altitudinal distance to treeline. We conclude that (1) the elevational response of tree-rings includes a threshold component (a minimal seasonal temperature) and that (2) radial growth is more strongly correlated with year to year variation in climate than with treeline elevation as such. Our data indicate that the current treeline position reflects influences of past climates and not the current climate.
Aim Here we examine the functional profile of regional tree species pools across the latitudinal distribution of Neotropical moist forests, and test trait–climate relationships among local communities. We expected opportunistic strategies (acquisitive traits, small seeds) to be overrepresented in species pools further from the equator, but also in terms of abundance in local communities in currently wetter, warmer and more seasonal climates. Location Neotropics. Time period Recent. Major taxa studied Trees. Methods We obtained abundance data from 471 plots across nine Neotropical regions, including c. 100,000 trees of 3,417 species, in addition to six functional traits. We compared occurrence-based trait distributions among regional species pools, and evaluated single trait–climate relationships across local communities using community abundance-weighted means (CWMs). Multivariate trait–climate relationships were assessed by a double-constrained correspondence analysis that tests both how CWMs relate to climate and how species distributions, parameterized by niche centroids in climate space, relate to their traits. Results Regional species pools were undistinguished in functional terms, but opportunistic strategies dominated local communities further from the equator, particularly in the Northern Hemisphere. Climate explained up to 57% of the variation in CWM traits, with increasing prevalence of lower-statured, light-wooded and softer-leaved species bearing smaller seeds in more seasonal, wetter and warmer climates. Species distributions were significantly but weakly related to functional traits. Main conclusions Neotropical moist forest regions share similar sets of functional strategies, from which local assembly processes, driven by current climatic conditions, select for species with different functional strategies. We can thus expect functional responses to climate change driven by changes in relative abundances of species already present regionally. Particularly, equatorial forests holding the most conservative traits and large seeds are likely to experience the most severe changes if climate change triggers the proliferation of opportunistic tree species.