# Summary paragraph outline (199 /  ideally of about 200 words, but certainly no more than 300 words)

Competition is very important to understand and predict the dynamics
of plant community composition. In terrestrial
vegetation, where plants strongly modify the environment in their
immediate neighbourhood, competition is conspusious but our ability to predict its consequences on plant performances is extremely limited. Predicting competition via phenotypic traits may be
simpler and more general than via competition coefficients between
each pair of species, an intractable approach at global scale. Here
for the first time we show how three functional traits - wood density,
specific leaf area and maximum height - have globally consistent
influences on growth and competition. Data are from forest inventories
monitoring growth for millions of trees throughout the world. Traits
had strong effects on maximum growth rates of species. Traits of focal
species also influenced their competitive response. Traits of
competitor species influenced their competitive effect. Smaller than
these effects, but still detectable, was a reduction in competition
when traits were less similar between the focal plant and its
competitors. Trait values that conferred stronger tolerance to
competition generally also rendered species slower growing in absence
of competition. This is an important trade-off to demonstrate at
global scale because it is a classical hypothesis for successional
coexistence of species in forest ecosystems.

# Main text (MAX 1500 words till the end of Main text plus summary paragraph = 1241)

Individuals interact in a myriad of different ways in ecological
communities. These interactions are crucial to understand and predict species composition and its
dynamics over time. Each individual modifies their immediate
environment and thus influences the performance of neighbouring
individuals. Negative influences are called competition and positive
influences  facilitation[@keddy_competition_2001].  Competition is
especially important for vegetation on land because in most vegetation
types depletion of light, water and nutrients is substantial. There
have been many studies on competition among
plants[@goldberg_patterns_1992; @keddy_competition_2001]. These traditionally have described competition via interaction coefficients between pairs of species. But this can quickly become intractable: the number of different interactions rises as $N^2$ with the number of species
$N$. Also this species-pair approach does not lead naturally to
generalization across different vegetation types and different
continents. Modeling
competition via phenotypic traits rather than via species might overcome these limitations and allow general
relationships to be established at at biome to global scale. However,
available
studies[@uriarte_trait_2010; @kunstler_competitive_2012; @hillerislambers_rethinking_2012; @lasky_trait-mediated_2014; @kraft_plant_2015] are too few and too local to allow broad generalization about how traits influence competition. Notably there is continuing debate about the relative importance of mechanisms whereby particular trait values confer competitive advantage, vs. mechanisms whereby competition is weaker when two species have dissimilar traits[@mayfield_opposing_2010]. This distinction is fundamental because if competition is driven mainly by trait similarity, this will favour coexistence of a wide spread of traits values.

Here we quantify competition as the influence of neighbours on growth of a focal tree. Our framework is novel in two important ways: (i) competition is analysed as a function of traits rather than of species at an unprecedented scale covering all the major biomes on Earth (Fig. \ref{ilustr}a) and (ii) the influence
of traits on competition is partitioned among four pathways (Fig. \ref{ilustr}b,c) as
follows. A competitive advantage for some trait values compared to
others can arise (1) through permitting faster maximum growth in
absence of competition[@wright_functional_2010]; (2) through better
competitive response [@goldberg_competitive_1996], growth of species
possessing those traits being less affected by competition from
neighbours; or (3) through stronger competitive effect[@goldberg_competitive_1996; @gaudet_comparative_1988], species
possessing those traits reducing more strongly the growth of their
neighbours because they are correlated with
stronger competitive effect. Finally (4) competition can promote trait
diversification, if decreasing trait similarity between the
competitors and the focal tree weakens competitive
interaction[@macarthur_limiting_1967]. These four pathways are likely
to be connected to the key traits used to describe plant
strategies[@uriarte_trait_2010; @kunstler_competitive_2012; @hillerislambers_rethinking_2012; @lasky_trait-mediated_2014; @kraft_plant_2015; @westoby_plant_2002; @chave_towards_2009],
and here we dissect how wood density, specific leaf area (SLA), and maximum height affect competition among neighbouring trees[@uriarte_neighborhood_2004]. We compiled data covering all the major biomes on Earth (Fig. \ref{res2}b) for basal area growth (increase in the area of the cross section of tree trunk at 1.3 m height) of more than 3 million trees representing more than 2500 species and extracted species means traits values from either the global TRY data base[@kattge_try_2011; @niinemets_global-scale_2001] or local
data bases (see Methods). We analysed how maximum growth of each
individual tree was reduced by the local density of competitors in its
neighborhood (measured as the sum of basal areas of competitors in
$m^2/ha$), accounting for traits of both the focal tree and its
competitors. This analysis allowed effect sizes to be estimated for
each of the four pathways (Fig. \ref{ilustr}c).

Across all biomes the strongest drivers of individual growth was, as
expected, a
positive effect of tree size (stem diameter). A negative effect of
the local abundance of competitors independent of their traits was the second strongest effect indicating that competition was dominant rather than facilitation. Then among trait
influences the most important were processes giving a competitive
advantage for some trait values compared to others. Strongest were
direct influences of traits on the focal plant’s growth in absence of
competition (Fig. \ref{res1} Extended data Table 3), with the fastest
growing species having low wood density and high SLA
(Fig. \ref{res1}). Then secondly some trait values led species to have
stronger competitive effect, or better competitive response
(Fig. \ref{res1}). Taken together these two processes were in the same
order as trait effects directly on maximum growth (Extended data Table
3; Fig \ref{res1}). Finally, a small but consistent effect showed that
higher trait similarity between focal and neighbour species resulted
in stronger competitive suppression of growth (Fig. \ref{res1}). This
process is capable of promoting trait diversity at a local scale. It
has generally been considered as the key mechanism by which traits
affect competition, but has rarely been quantified with field
data[@mayfield_opposing_2010]. Our analysis shows that at global scale
this process is present but not dominant. Analyses allowing for
different effects among biomes did not show strong evidence for any
particular biome behaving consistently differently from the others
(Fig. \ref{res3}). Results varied most among biomes for SLA, and this
may reflect fundamental differences between deciduous and evergreen
plant strategies[@lusk_why_2008] (futher detail in Supplementary discussion).

Importantly, our study also supported the idea that trait values
favouring high tolerance to competition or high competitive effect also render species slow growing in absence
of competition, and that this trait-based trade-off is a consistent
global phenomenon common to all forested biomes. This trade-off
(Fig. \ref{res3}) is significant because it is a classical explanation
for successional coexistence of species in
forests[@rees_long-term_2001]. Although confidence intervals were
wide, trade-off was present for all three traits (crossover between
high and low trait values in \ref{res3}), and the directions of trait-effects underpinning them agreed well with existing literature (further detail in Supplementary Discussion). High wood density was associated with slow potential growth rate but high tolerance to competition (Fig. \ref{res1}), in
agreement with shade-tolerant species having high wood
density[@wright_functional_2010]. High SLA
was correlated with a fast maximum growth, as reported in previous
studies on trees[@wright_functional_2010], but with a weakened
competitive effect, in agreement with the observation that the shorter
leaf life span associated with high SLA results in low light
interception[@niinemets_review_2010]. Tall maximum height was
positively related to maximum growth in most biomes, as previously
reported[@wright_functional_2010] (though with wide confidence
interval in all biomes expect temperate rainforest), but
with a lower tolerance to competition than shorter species, in line
with the idea that sub-canopy trees are more
shade-tolerant[@poorter_architecture_2006]. Coordination between trait values conferring high competitive effect and trait values conferring high competitive tolerance has been widely expected[@goldberg_competitive_1996; @kunstler_competitive_2012]. However, in agreement with previous studies[@goldberg_components_1990; @goldberg_competitive_1991; @wang_are_2010], we found little evidence for such coordination. It was present only for wood density, where high density conferred better competitive response and also stronger competitive effect (but with wide confidence interval, Fig. \ref{res2}). Finally, the underlying mechanisms that may explain the trait similarity effects are unknown for these traits, but could include neighbouring species with similar traits supporting heavier loads of specialised pathogens[@bagchi_pathogens_2014], capturing light less efficiently[@sapijanskas_tropical_2014] or recycling litter less efficiently[@sapijanskas_beyond_2013].

The globally consistent links that we report here between traits and
competition have considerable promise for predicting
complex species interactions governing forest
communities at a global scale. A challenge for the future is to analyse survival and
recruitment as well as growth on the basis of traits rather than of
species. This would span the whole of life-history and build a more
complete picture of the fitness consequences of different
traits. These results on growth already demonstrate, in agreement with
previous studies[@mayfield_opposing_2010; @adler_trait-based_2013; @kraft_plant_2015],
that trait dissimilarity is not the major determinant of local scale
competition among plants. However at larger scale, a mosaic of
successional stages can favour different compromises along a trade-off
between potential maximum growth and performance in condition of high competition, supporting a diversity of traits in this way.


**Supplementary Information** is available in the online version of the paper.

**Acknowledgements**
We are grateful to people whose long term commitment established and maintained the forest plots and their associated databases, and who granted us access  - New Zealand, Japan, MAGRAMA for the Spanish Forest Inventory, France, Switzerland, Sweden, US, Canada (for the following provinces: Quebec, Ontario, Saskatchewan, Manitoba, New Brunswick, and Newfoundland and Labrador), CTFS plots (BCI, Fushan and Luquillo) and Cirad permanent plots (Paracou, M’Baïki). GK was supported by a Marie Curie International Outgoing Fellowship within the 7th European Community Framework Program (Demo-traits project, no. 299340). The working group that initiated this synthesis was supported by Macquarie U and by Australian Research Council through a fellowship to MW.



**Author contributions**
GK conceived the study and lead a workshop to develop this analysis with the participation of DAC, DF, FH, RMK, DCL, LP, MV, GV, JSW, and MW. GK wrote the manuscript with input from all authors. G.K. devised the main analytical approach, wrote the computer code and performed analyses. G.K. and F.H. processed the data. GK, DAC, DF, FH, RMK, DCL, MV, GV, JSW, MA, CB, JC, JHCC, SGF, MH, BH, JK, HK, YO, JP, HP, MU, SR, PRB, IFS, GS, NS, JT, BW, CW, MAZ, HZ, JZ, NEZ collected and processed the raw data.

**Author information**

\newpage

# FIGURES & TABLES (legend total 483 /500  max words)

![**Assessing competitive interactions at global scale.** **a,** Precipitation-temperature space occupied by each data set (NFI data : national forest inventories, LPP data : large permanent plots). For data with multiple plots, the range of climate for each data is represented by an ellipse covering 98% of the plots. Biomes definition from @ricklefs_economy_2001: 1, tundra; 2, taiga; 3, mediterranean; 4, temperate forest; 5, temperate rainforest; 6, desert; 7, tropical seasonal forest; 8, tropical rainforest. **b,** Sampled patches vary in the density of competitors from species $c$ around individuals from a focal species $f$. **c,** We modeled how trait ($t_f$) of the focal tree, and the density (measured as the sum of their basal area - area of the cross section of tree trunk at 1.3 m height) and traits ($t_c$) of competitors species influence basal area growth of the focal tree. Maximum growth (red) is influenced by trait of the focal tree. Reduction in growth per unit basal area of competitor (blue, $\alpha_{c,f}$) is modelled as growth reduction independent of trait ($\alpha_0$), the effect of the focal tree’s traits on its competitive response ($\alpha_r \, t_f$), the effect of competitor traits on their competitive effect ($\alpha_e \, t_c$), and the effect of trait similarity between the focal tree and its competitors on competition ($\alpha_s \, \vert t_c-t_f \vert$). The parameters $m_1, \gamma, \alpha_0, \alpha_e, \alpha_r$ and $\alpha_s$ are fitted from data. \label{ilustr}](image/fig1f.pdf)


\newpage


![**Trait effects and trait-independent effects on maximum growth and competition.** Standardized regression coefficients for growth modeled, fitted separately for each trait (points: mean estimates and lines: 95% confidence intervals). The parameter estimate represents: maximum growth variation with trait of the focal tree $m_1$, the competition effect independent of traits $\alpha_0$, the effect of the focal tree’s traits on its competitive response $\alpha_r$ (positive $\alpha_r$ indicates that high trait values of the focal tree limits its growth reduction via competition), the effect of competitor traits on their competitive effect $\alpha_e$ (negative $\alpha_e$ indicates that high trait values of the competitors increase their competitive reduction of the growth of the focal tree), and the effect on competition of trait similarity between the focal tree and its competitors $\alpha_s$ (negative $\alpha_s$ indicates that high trait similarity worsen the growth reduction). \label{res1}](../../figs/figres1.pdf)

\newpage


![**Variation between biomes of trait effects on maximum growth and competition.** Standardized regression coefficients of the growth models fitted separately for each trait as in Fig. 2, but with separate estimates for each biome (see Fig 1a. for definitions of the biomes). Tropical rainforest and tropical seasonal forest were merged together as tropical forest, tundra was merged with taiga, and desert was not included as too few data were available. \label{res2}](../../figs/figres2.pdf)

\newpage

![**Trade off between growth without competition and growth with competition underpinned by maximum height, specific leaf area, and wood density.** The growth of a focal tree with low or high trait value (respectively 5 and 95% quantile) in function of the local basal area of competitors with trait value resulting in a high competitive impact. The shaded area represents the 95% confidence interval of the prediction. \label{res3}](../../figs/figres3.pdf)



# References (max 30)