Biogeosciences, 9, 775–801, 2012 www.biogeosciences.net/9/775/2012/ doi:10.5194/bg-9-775-2012 © Author(s) 2012. CC Attribution 3.0 License.

Biogeosciences

Coordination of physiological and structural traits in Amazon forest trees ˜ 1,2,† , N. M. Fyllas2 , T. R. Baker2 , R. Paiva3 , C. A. Quesada2-4 , A. J. B. Santos3,4,† , M. Schwarz1 , S. Patino H. ter Steege5 , O. L. Phillips2 , and J. Lloyd2,6 1 Max-Planck-Institut

f¨ur Biogeochemie, Postfach 100164, 07701, Jena, Germany of Geography, University of Leeds, LS2 9JT UK 3 Institito Nacional de Pesquisas da Amazˆ onia, Manaus, AM, Brazil 4 Departamento de Ecologia, Universidade de Bras´ılia, DF, Brazil 5 Dept. of Plant Ecology and Biodiversity, Utrecht University, The Netherlands 6 School of Earth and Environmental Sciences, James Cook University, Cairns, Qld 4871, Australia † deceased 2 School

Correspondence to: J. Lloyd ([email protected]) Received: 5 May 2011 – Published in Biogeosciences Discuss.: 25 May 2011 Revised: 16 November 2011 – Accepted: 18 January 2012 – Published: 16 February 2012

Abstract. Many plant traits covary in a non-random manner reflecting interdependencies associated with “ecological strategy” dimensions. To understand how plants integrate their structural and physiological investments, data on leaf and leaflet size and the ratio of leaf area to sapwood area (8LS ) obtained for 1020 individual trees (encompassing 661 species) located in 52 tropical forest plots across the Amazon Basin were incorporated into an analysis utilising existing data on species maximum height (Hmax ), seed size, leaf mass per unit area (MA ), foliar nutrients and δ 13 C, and branch xylem density (ρx ). Utilising a common principal components approach allowing eigenvalues to vary between two soil fertility dependent species groups, five taxonomically controlled trait dimensions were identified. The first involves primarily cations, foliar carbon and MA and is associated with differences in foliar construction costs. The second relates to some components of the classic “leaf economic spectrum”, but with increased individual leaf areas and a higher 8LS newly identified components for tropical tree species. The third relates primarily to increasing Hmax and hence variations in light acquisition strategy involving greater MA , reductions in 8LS and less negative δ 13 C. Although these first three dimensions were more important for species from high fertility sites the final two dimensions were more important for low fertility species and were associated with variations linked to reproductive and shade tolerance strategies. Environmental conditions influenced structural traits with ρx of individual species decreasing with increased soil fertility and higher temperatures. This soil fertility response

appears to be synchronised with increases in foliar nutrient concentrations and reductions in foliar [C]. Leaf and leaflet area and 8LS were less responsive to the environment than ρx . Thus, although genetically determined foliar traits such as those associated with leaf construction costs coordinate independently of structural characteristics such as maximum height, others such as the classical “leaf economic spectrum” covary with structural traits such as leaf size and 8LS . Coordinated structural and physiological adaptions are also associated with light acquisition/shade tolerance strategies with several traits such as MA and [C] being significant components of more than one ecological strategy dimension. This is argued to be a consequence of a range of different potential underlying causes for any observed variation in such “ambiguous” traits. Environmental effects on structural and physiological characteristics are also coordinated but in a different way to the gamut of linkages associated with genotypic differences.

1

Introduction

Plant traits are widely used in ecology and biogeochemistry. In particular, sets of functional characters can serve as the basis for identifying important adaptations that improve the success of different taxa at different environments. Over the last decade significant advances have been made in terms of our understanding of plant trait inter-relationships and associated trade-offs (Reich et al., 1997; Westoby et al., 2002),

Published by Copernicus Publications on behalf of the European Geosciences Union.

˜ et al.: Tropical tree trait dimensions S. Patino

776 especially in terms of the so called “leaf economic spectrum” (Wright et al., 2004) with well documented systematic and co-ordinated changes in leaf nitrogen and phosphorus concentrations, leaf mass per unit area, MA and leaf lifetimes. Attention has also been paid to the relationships between physiological and structural characteristics of leaves and other plant traits. For example, it has been reported that leaf size declines with wood density, ρw (Pickup et al., 2005; Wright et al., 2006, 2007; Malhado et al., 2009) and it has been suggested that this is because the ratio of leaf area to sapwood area (8LS ) should also decline with increasing wood density due to hydraulic constraints (Wright et al., 2007). Nevertheless, although 8LS may decline with ρw for trees in some ecosystems that are clearly waterlimited (Ackerly, 2004; Cavender-Bares et al., 2004), 8LS sometimes actually increases with ρw (Wright et al., 2006; Meinzer et al., 2008). The latter study also found that associated with these higher 8LS and high wood density stems were lower stem hydraulic conductances, more negative midday leaf water potentials, and more negative bulk leaf osmotic potentials at zero turgor. Thus, leaves of some high wood density species may be characterised by physiological and structural adaptations allowing them to function at more severe water deficits than is the case for low wood density species. The Panama study of Meinzer et al. (2008) also found that higher ρw species tended to have higher MA . Although similar positive correlations between MA and ρw have also been reported for other ecosystems (e.g. for sclerophyllous forest: Ishida et al., 2008) when examining the bivariate relationship between ρw and MA across a range of tropical forest sites, Wright et al. (2007) observed no significant relationship. Likewise, when examining variation in leaf and stem traits for 17 dipterocarp species growing in a common garden in southern China, Zhang and Cao (2009) also found no significant correlation between ρw and MA . Variations in MA may also be related to a suite of additional plant physiological characteristics (Poorter et al., 2009), varying negatively with dry-weight foliar nitrogen and phosphorus concentrations (Wright et al., 2004; Fyllas et al., 2009) as well as tending to increase with increasing tree height (Thomas and Bazzaz, 1999; Kenzo et al., 2006; Lloyd et al., 2010). Potential tree height, Hmax , has also been related to a number of wood traits (Chave et al., 2009) with taller plants tending to have bigger conduits in their trunks, but fewer conduits overall (Coomes et al., 2007). Within a given stand, taller and generally more lightdemanding rain forest species also tend to have larger leaves, this being associated with shallower crown and a more efficient light capture (Poorter et al., 2006; Poorter and Rozendaal, 2008). Leaf–size may also be influenced by other factors. For example, Australian rain forests growing on oligotrophic soils typically have a greater abundance of smaller leaved species than for nearby forests found on more mesotrophic soil types (Webb, 1968). Biogeosciences, 9, 775–801, 2012

Seed size may also relate to the above plant functional traits. For example, one of “Corner’s rules” describes a tendency for species with thick twigs to have large appendages (leaves and fruit). The range of viable seed size also tends to increase with plant height (Moles et al., 2005; Grubb et al., 2005). Forests on the more fertile soils of western Amazonia tend to have smaller average seed masses than their less fertile counterparts on the Guyana Shield and elsewhere (ter Steege et al., 2006), this perhaps being related to several advantages attributable to large seeded species under nutrientpoor conditions, viz. greater initial nutrient stores, greater initial root zone expansion, and increased mychorrizal infection, all of which would be expected to increase the probability of seedling survival (Foster, 1986). This paper presents new data on leaf and leaflet size and 8LS for 661 species located in 52 plots across the Amazon Basin. The trees sampled form a subset of those also examined for variations in branch xylem density (Pati˜no et al., 2009), and for foliar nutrients, MA and δ 13 C (Fyllas et al., 2009), which had previously been analysed separately. We here investigate the inter-relationships between these structural and physiological parameters also considering taxonomic variations in Hmax (Baker et al., 2009) and seed mass (ter Steege and Hammond, 2001; ter Steege et al., 2006). Specifically, we were interested to assess the degree to which the observed variations in the studied structural and physiological traits were coordinated with each other into identifiable integrated trait dimensions: for example, those associated with leaf construction costs, light acquisition, and/or shade tolerance. 2 2.1

Materials and methods Study sites

In the analysis here, RAINFOR sample plots have been aggregated as discussed in Fyllas et al. (2009), with further plot details available in Pati˜no et al. (2009) and Quesada et al. (2010). Ten plots in Fyllas et al. (2009) have not been included due to insufficient structural trait data having been collected, but the range of soils encountered here is still substantial with the sum of exchangeable bases (0– 0.3 m), for example ranging from less than 1 mmolc kg−1 to nearly 100 mmolc kg−1 . Total soil phosphorus ranged from 26 mg kg−1 for an ortseinc podzol to 727 mg kg−1 for a eutric cambisol (Quesada et al., 2010). Mean annual precipitation varies from less than 1.5 m a−1 on sites at the north and southern periphery of the basin to more than 3.0 m a−1 for sub-montane sites close to the Andes.

www.biogeosciences.net/9/775/2012/

˜ et al.: Tropical tree trait dimensions S. Patino 2.2

777

Structural traits

For most trees sampled in Pati˜no et al. (2009) and Fyllas et al. (2009), and from the same terminal branches for which data has already been presented in those studies, all leaves from the branch had also been counted. From that branch, a sub-sample of 10–20 leaves was randomly chosen to estimate individual leaf area, LA , and leaflet area, `A (when a species had compound leaves), and to estimate the total leaf area of the branch. All age and size leaves or leaflets were selected for this analysis except for very young leaves or those which were obviously senescent. The chosen leaves were usually scanned fresh on the same day of collection. When this was not possible the same day, they were stored for a maximum of two days in sealed plastic bags to avoid desiccation and any consequent reduction of the leaf area. Scans were analysed using “Win Folia Basic 2001a” (Regent Instruments Inc., 4040 rue Blain Quebec, QC., G2B 5C3 Canada) to obtain LA and `A . The distal (sapwood + pith) and pith diameters for each branch were also measured with a digital caliper (Mitutoyo Corporation, Japan) with sapwood area, AS , then estimated by subtracting pith area from the total branch area with 8LS =nL¯ A /AS where n is the number of leaves distal to the piece of branch sampled and L¯ A is the average area of the individual leaves sub-sampled for the estimation of LA and/or `A . Branch xylem density data for the same samples were obtained as described in Pati˜no et al. (2009). In brief, this consisted of the estimation of the volume of a branch segment, approximately 1 cm in diameter and 5–10 cm long using calipers, with the pith removed as necessary and dry weight subsequently determined. Species maximum height taken from the database developed by Baker et al. (2009) with estimates made to the species level for 80% of the trees identified, and the bulk of the remainder being genus level averages. Seed mass (S) was taken as a genus level dependent variable and was already on a log10 ordinal scale (ter Steege et al., 2006). 2.3

Physiological foliar traits

Foliar traits used here are as described/measured in Fyllas et al. (2009) and Lloyd et al. (2010) and include leaf mass per unit area (MA ) and foliar [N], [C], [P], [Ca], [K] and [Mg] expressed on dry-weight basis. Foliar 13 C/12 C discrimination, 1, was estimated from measurements of foliar δ 13 C (Fyllas et al., 2009) using an assumed value for the isotopic composition of source air equal to −8.0 ‰ (Farquhar et al., 1989) and subsequently transformed to a diffusional limitation index, , according to (Fyllas et al., 2012) r  = 1−

(1 − 4.4)/25.6 − 0.2 0.8

www.biogeosciences.net/9/775/2012/

which utilises the well known relationship between 1 and the ratio of internal to ambient CO2 concentrations, ci /ca (Farquhar et al., 1989). Equation (1) assumes that at current day ca , photosynthesis can be considered a roughly linear function of ci and with a maximum practical ci /ca (indicating minimal diffusional limitation) of 0.8. Here we have taken a value of 4.4 ‰ for the fractionation against 13 CO2 during diffusion into the leaf and 30.0 ‰ for the fractionation against 13 CO2 during photosynthetic fixation (Farquhar et al., 1989). Increasing  values are associated with lower ci /ca , and thus, other things being equal, a higher water use efficiency, W , this being the ratio of carbon gained to water lost during photosynthetic CO2 assimilation. Equation (1) relies on a simplified expression for 1 which ignores difference between gas- and liquid-phase fractionations within the leaf (Farquhar et al., 1989), but this should not seriously compromise its utility in the current context. 2.4

Climate and soils

The soil and climate predictors table used was the same as in Fyllas et al. (2009), using a set of measured soil properties (Quesada et al., 2010) with precipitation variables and temperature from the “WorldClim” dataset (http://www. worldclim.org). Estimates of mean annual solar radiation are from New et al. (2002). As in Fyllas et al. (2009) we separate soils into two fertility classes based on their total phosphorus concentration and the total sum of reserve bases, (Quesada et al., 2010). In brief this categorisation gives rise to arenosols, podzols, ferralsols, and most acrisols being classified as low fertility soils. High fertility soils include plinthosols, cambisols, fluvisols, gleysols and most alisols. 2.5

Statistical analysis

This paper implements a similar set of statistical analyses to that described in detail in Fyllas et al. (2009). Preliminary tests included analysis of normality (Shapiro-Wilk) and homogeneity of variance (Fligner-Killeen) for each of the structural traits of interest. The foliar related structural traits (LA , `A and 8LS ) presented a right skewed distribution and thus were all log10 transformed. As ρx , Hmax and S (the latter already provided as size classes on a log10 scale) were more or less symmetrically distributed around their mean we did not apply this transformation for these variables, even though the Shapiro test failed to identify strict normality. The nonparametric Kruskal-Wallis test (Hollander and Wolfe, 1999) was used to explore for differences between fertility groups as well as for differences between families, genera within a family and species within a genus. All analyses were performed with the R statistical platform (R Development Core Team, 2010).

(1) Biogeosciences, 9, 775–801, 2012

˜ et al.: Tropical tree trait dimensions S. Patino

778 2.5.1

Partitioning of variance and estimation of taxonomic and environmental effects

A multilevel model was initially fitted for all traits (including those previously analysed separately in Fyllas et al. (2009) and Pati˜no et al. (2009) because this was a slightly different dataset), except Hmax and S according to 2 = µ + p + f/g/s + ,

(2)

where µ is the overall mean value of each trait, 2; p is the plot effect, i.e. the effect of the location that each individual is found, and f/g/s represents the genetic structure of the data, i.e. that each individual belongs to a species (s), nested in a genus (g), nested in a family (f ), and  is the error term. All parameters were estimated by the Residual Maximum Likelihood (REML) method with the lme4 library available within R (Bates and Sarkor, 2007). Fyllas et al. (2009) have already discussed further details of the above formulations and the advantage in being able to partition the variance from the family to the species level, also taking into account the location (thus the environmental contribution to trait variation) where the trait was measured. The Supplementary Information (II) of that paper also provides an empirical validation of the approach used. Note, that whilst theoretically possible, we do not include interaction terms in Eq. (2), this is because there is insufficient species replication across different sites. Nevertheless, investigations into the likely magnitude of such effects have been undertaken as part of the analyses in both Pati˜no et al. (2009) and Fyllas et al. (2009) and have not been found to be significant. Again we were interested in exploring the taxonomic (estimated as the sum of family ± genus ± species random effects) and environmental terms, using bivariate relationships as well as multiple nonparametric regressions of plot effect contributions on a set of environmental predictors. For the latter we used Kendall’s τ as our measure of association calculating the significance of partial correlations using our own specifically written code, using the R statistical platform. For Hmax and S no multilevel model was fitted or environmental effect assumed, the available data being considered to express directly the genetic potential of each species. We also note that our estimates of S are resolved at the genus level only (ter Steege and Hammond, 2001) and are only on a log10 categorical scale. This introduces potential errors into the analyses where S is involved because all other traits have been resolved at the species level. Thus, even though a small portion of the observed variation in S generally occurs at the species level (Casper et al., 1992), bivariate and multivariate analyses involving this trait as presented here may carry somewhat more “noise” than would otherwise be the case. 2.5.2

Bivariate relationships

Relationships were initially assessed with the Pearson’s correlation coefficient (r) with subsequent Standardized MaBiogeosciences, 9, 775–801, 2012

jor Axis (SMA) line fits where significant correlations were identified. In this study, SMA line fits are applied to the raw dataset (including all measured traits and thus intraspecific variation), to the taxonomic component of trait variation (i.e. each species is represented by a single data point) as well as to the plot level effects (i.e. each plot is contributing a single data point). In each case we initially fitted separate lines for each fertility group, and when a common SMA slope was identified we tested for differences in elevation and/or slope between fertility groups, using the smartr library available within R (Warton et al., 2006). We explored the plot level effect of each structural trait, through non-parametric correlation analysis on selected soil and environmental predictors, with the soil variables reduced to three principal axes to avoid multicollinearity (Fyllas et al., 2009). The climatic variables of mean annual temperature, total annual precipitation, dry season precipitation and mean annual radiation were also examined. As extensively discussed in Fyllas et al. (2009) we dealt with spatial autocorrelation issues by fitting appropriate simultaneous autoregressive models (SAR) which include a spatial error term (Lichstein et al., 2002) to help interpret the significance of full and partial Kendall’s τ coefficients as a measure of association between plot-level trait effects and environmental predictors. 2.5.3

Multivariate analyses

Inferred taxonomic effects were analysed jointly for species found on fertile versus infertile soils (excluding those found on both soil types) by calculating separate variance– covariance matrices for the two species groups and then using the common principal components (CPC) model of Flury (1988) as implemented by Phillips and Arnold (1999). Within this model, it is assumed that the two populations of species have the same eigenvectors (principal components; denoted here as U ) but that the relative loading of the various U as expressed through their eigenvalues (λ) may potentially vary between the two populations. Flury’s model provides a hierarchy of tests corresponding to a range of possible relationships between matrices including equality, proportionality, common principal components, partial common principal components or unrelated (Flury, 1988; Phillips and Arnold, 1999). CPC can thus be seen as a method for summarizing the variation in two or more matrices. Nevertheless, caution needs to be applied when using CPC to address the more complex goal of diagnosing and understanding the nature of the changes that underlie the difference between the matrices. This is because CPC tends to spread any differences over many of the vectors it extracts and often over all of them (Houle et al., 2002). As the CPC model does not strictly apply to correlation matrices (Flury, 1988), we standardised each variable before calculating the input variance–covariance matrix by dividing each variable by its observed range (across both high and www.biogeosciences.net/9/775/2012/

779

Population density

Population density

˜ et al.: Tropical tree trait dimensions S. Patino

log10[leaf area] (m2) Population density Population density

log10[leaf mass per unit area] ( g m-2) log10[leaf area:sapwood area] ( m-2 cm-2)

Population density

Population density

***

log10[leaflet area] (m2)

Diffusional limitation index

Population density

Population density

Branch xylem density (kg m-3)

Maximum species height (m)

-6

-4

-2

0

2

log10[seed mass] (g)

Fig. 1. Probability density histograms of raw data per fertility group for leaf area (LA ; m2 ), leaflet area (`A ; m2 ), leaf mass per unit area (MA ; g m−2 ), (m2 ), leaf area:sapwood area ratio (8LS ; cm2 m−2 ), branch xylem density (ρx ; kg m−3 ),  = stomatal limitation index (dimensionless; see Eq. 1), species maximum height (Hmax ; m) and seed mass (S; g). Open red bars represent low and blue dashed bars high soil fertility plots, as defined by the quantitative determinations of the level of total reserve bases from 0.0–0.3 m depth (Fyllas et al., 2009; Quesada et al., 2010). Also given for each histogram are the mean and the variance for each trait. Significant differences in mean values and/or variances between the two fertility groups were identified with the Fligner-Killeen test respectively. Significance codes: *** < 0.001, ** < 0.01,* < 0.05.

Fig. 1. Probability density histograms of raw data per fertility group for leaf area (LA ; m2 ), leaflet area (`A ; m2 ), leaf mass per unit area −2 2 −2 (MAsoils) ; gm ), (mby2 ), leaf area:sapwood areamultivariate ratio (Φ cmPCA mof the),derived low fertility as first proposed Gower (1966) but, due All other analyses LS ; (e.g. to the presence of the occasional outlier, taking the effective −3 environmental effects) were implemented with the ade4 branch xylem (ρxof; thekgU and m ),  =(Thioulouse stomatal index range as the 0.1 to 0.9 quantiles.density Standard errors package et al.,limitation 1997) available within the R staλ for the CPC models were estimated assuming asymptotic tistical platform with the environmental effect PCA (dimensionless; see Eq. 1), species maximum height (Hmax ; m) and undernormality as described in Flury (1988). taken on the correlation matrix. seed mass (S; g). Open red bars represent low and blue dashed bars high soil fertility plots, as defined by the quantitativeBiogeosciences, determinations www.biogeosciences.net/9/775/2012/ 9, 775–801, 2012 of the level of total reserve bases from 0.0–0.3 m depth (Fyllas et

˜ et al.: Tropical tree trait dimensions S. Patino

780 3 3.1

Results Trait distribution in relation to soil type

The structural traits distributions along with those for MA and  for the complete dataset divided to low and high fertility groups are shown in Fig. 1 with overall mean values, range and variances for each plot for all traits also provided in the Supplementary Information (Table S1). The three leaf related traits introduced here (LA , `A and 8LS ) did not differ significantly between low and high fertility sites (Fig. 1). On the other hand, ρx and S showed significant differences between the two fertility groups, with their distributions shifted to the left for fertile sites, i.e. higher ρx and S were found for species found on infertile soils. This is similar to the shifted distributions identified for most leaf mineral concentrations across fertility gradients (Fyllas et al., 2009) but in the opposite direction, i.e. with higher structural carbon and lower mineral investment in less fertile environments. As expected from our prior analysis of the statistical distribution of foliar δ 13 C (Fyllas et al., 2009), the diffusional limitation index of Eq. 1  tended to be lower for trees growing on low fertility soils. Despite a difference in variance between low and high fertility sites, there was, however, no overall effect of soil fertility classification on the average Hmax . 3.2

Partitioning of the variance

The variation apportioned to different taxonomic levels varies for each of the traits examined (Fig. 2). When leaf size was expressed per leaflet, most of the variation was attributed at the species level (0.31) with the overall taxonomic component (i.e. family ± genus ± species) adding up to a very high (0.62) proportion. When leaf size was expressed at the leaf level, most of the variation was attributed at the family level (0.29) with a very high overall taxonomic component (0.71). In contrast to LA and `A , plot level contributions to the total variance were substantial for the other structural traits: being around 0.30 for ρx and 0.27 for 8LS . These are not necessarily higher than their respective taxonomic components, but underline the importance of the site growing conditions in influencing structural traits such as ρx and 8LS . As for the foliar traits reported in Fyllas et al. (2009) this must have direct implications for different physiological processes. In that study, leaf mass per unit area and [C], [N] and [Mg] emerged as highly constrained by the taxonomic affiliation, but with others, such as [P], [K] and [Ca] also strongly influenced by site growing conditions. That study also found foliar δ 13 C to be strongly influenced by site growing conditions, consistent with its analogue here () having its environmental component as the dominant source for its variation. Overall, there was a tendency for the residual component (related to intraspecies variations not accountable for by different plot locations and experimental error) to increase as the proportion of variation accountable for by taxonomic affiliation declined Biogeosciences, 9, 775–801, 2012

Diffusional limitation index

Leaf area: sapwood area ratio

Branch xylem density

Leaf area

Leaflet area

Proportion of total variance

Fig. 2. Partitioning of the total variance for each studied property into taxonomic (family/genus/species), environmental (plot) and erFig. 2. Partitioning of the total variance for each structural property ror (residual) components. Traits are sorted from less to more taxinto taxonomic (family/genus/species), environmental (plot) and an onomically constrained. Significance of each variance component error (residual) components. Foliar properties are sorted from less was tested with a likelihood ratio test (Galwey, 2006). Significance to more taxonomically constrained. Significance of each variance codes: *** < 0.001, < 0.01,* < 0.05.ratio test (Galwey, 2006). component was tested**with a likelihood Significance codes: *** < 0.001, ** < 0.01,* < 0.05.

and with the proportion attributable to plot location tending to increase as the residual component became larger. 3.3

Bivariate relationships: raw data

These are not considered in any detail here, but for the interested reader data are summarised in the Supplementary Information, Table S2A. 3.4

Bivariate relationships: taxonomic components

Considering data from both low and high fertility sites together, Table 1 lists correlations and SMA slopes for the derived taxonomic components with this same information shown in more detail (including confidence intervals) in the Supplementary Information (Table S2A) and with low and high fertility species separated for OLS and SMA regression analyses in Table S2B. Within Table 1, the SMA slopes reflect the relationship y ↔ x, with the x as the column headers and the y being the row labels. Figures 3 through 6 illustrate the more important relationships involving the sampled structural traits. Due to considerations associated with multiple testing, we focus only on relationships significant www.biogeosciences.net/9/775/2012/

˜ et al.: Tropical tree trait dimensions S. Patino

781

Table 1. Relationships between the derived genetic components of the observed plant traits: MA = leaf mass per unit area (gm−2 ); elemental concentrations are on a dry weight basis (mg g−1 ), LA = leaf area (m2 ), `A = leaflet area (m2 ), 8LS = leaf area:sapwood area ratio (cm2 g−1 ), ρx = branch xylem density (kg m−3 ),  = stomatal limitation index (see Eq. 1), S = seed mass (g), Hmax = species maximum height (m). Values above the diagonal represent the slope of the relationship (y axis as columns labels, x axis as row labels). Values below S. Pati˜no et al.: Tropical tree trait dimensions the diagonal represent the correlation coefficient. Values significant at P < 0.05 are given in bold. NS = no slope estimated as the relationship was not significant. [C]

log[N]

log[P]

log[Ca]

log[K]

log[Mg]

log(LA )

log(`A )

log(8LS )

ρx



log(S)

Hmax

− 0.15 −0.43 −0.41 −0.07 −0.28 −0.14 −0.09 0.17 −0.24 0.13 0.09 0.12 0.17

0.37 − 0.07 -0.02 −0.51 −0.45 −0.45 0.14 −0.08 0.06 0.07 0.03 0.18 0.04

−1.01 NS − 0.66 0.02 0.18 0.05 0.27 −0.11 0.20 −0.08 0.23 −0.16 −0.03

−1.21 NS 1.20 − 0.14 0.46 0.18 0.37 0.04 0.14 −0.20 0.27 −0.08 0.00

NS −6.28 NS 1.93 − 0.46 0.65 −0.03 0.07 0.00 −0.21 0.12 −0.34 −0.06

−1.65 −4.43 1.63 1.36 0.7 − 0.59 −0.01 0.18 0.04 −0.24 0.06 −0.23 −0.02

−2.18 −5.85 NS 1.79 0.93 1.32 − −0.14 0.15 −0.09 −0.12 0.09 −0.25 −0.08

NS 17.30 6.36 5.37 NS NS −2.98 − 0.41 0.26 −0.10 0.12 0.02 −0.02

4.32 NS −4.18 NS NS 2.60 1.97 0.65 − 0.01 −0.22 −0.08 0.00 0.00

−1.27 NS 1.22 1.03 NS NS NS 0.19 NS − 0.07 −0.08 −0.10 −0.07

0.88 NS NS −0.72 −0.37 −0.52 −0.40 −0.13 −0.2 NS − −0.09 0.25 −0.03

NS NS 0.22 0.19 0.10 NS 0.10 0.04 NS NS NS − −0.20 0.11

19.2 51.4 −18.6 NS −8.3 −11.4 −8.8 NS NS NS 21.6 −81.8 − 0.14

167 NS NS NS NS NS NS NS NS NS NS 731 8.8 −

Leaf mass per unit area (g m-2)

log(MA ) [C] log[N] log[P] log[Ca] log[K] log[Mg] log(LA ) log(`A ) log(8LS ) ρx  log(S) Hmax

MA

(a)

(b) Seed mass (mg)

Variable

Species maximum height (m) Fig. 3. Standard Major Axis (SMA) regression lines between species maximum height (Hmax ) and the derived taxonomic components of leaf mass per unit area (MA ) for the same species and the associated average seed mass (S) for the associated genus. Red open circles indicate species found on low fertility sites and the blue open circles indicate species found on high fertility sites. Species found on both soil fertility groups are indicated with closed circles (see text for details). Red solid lines show the SMA model fit for low fertility species which is significantly different to the blue solid lines for high fertility soil species.

Fig. 3. Standard Major Axis (SMA) regressions lines between species maximum height (Hmax ) and the derived taxonomic components of leaf mass per unit area (MA ) for theSMA sameslope species and intercept between the species associated at p ≤ 0.001 though, where interesting and/or informative, and/or the associated average seed mass (S) for the associated genus. Red statistically less significant relationships are also considered. with the two soil fertility classes (see Supplementary Inforopen circles indicate species found on low fertility sites and theS2B) blue we have fitted separate lines for species mation, Table open circles indicate species found on high fertility sites. Species found on low and high fertility soils. This shows that for 3.4.1 Maximum tree height found on both soil fertility groups are indicated species with closed circles with low fertility soils, both MA and S associated (see text for details). Red solid lines show the SMA model fit which higher at a given Hmax than their higher tend to be slightly Generally only poor correlations were observed for Hmax , is significantly different to the blue solid lines for high fertility soil fertility counterparts. Especially for S ↔ Hmax the variathese being significant only for log10 (MA ) (p ≤ 0.001) and species. tion is considerable, particularly at low Hmax , with S varying log10 (S) (p ≤ 0.01). The MA ↔ Hmax and S ↔ Hmax relathree orders of magnitude for Hmax between 10 and 30 m. tionships are shown in Fig. 3. Here, due to differences in the www.biogeosciences.net/9/775/2012/

Biogeosciences, 9, 775–801, 2012

˜ et al.: Tropical tree trait dimensions S. Patino

Leaf phosphorus (mg g-1)

(a)

(b)

(d)

(c) Seed mass(g)

Leaf potassium (mg g-1)

Leaf mass per unit area (g m-2)

782

Branch xylem density (kg m-3) Fig. 4. Standard Major Axis (SMA) regression lines between the derived species components of branch xylem density (ρx ) and those for mass per unit area (MA ), foliar [P] and foliar [K] for the same species and the average seed mass (S) for the associated genus. Red open circles indicate species found on low fertility sites and the blue open circles indicate species found on high fertility sites. Species found on both soil fertility groups are indicated with closed circles (see text for details). The black solid lines show the SMA model fit which did not depend on soil fertility.

Fig. 4. Standard Major Axis (SMA) regressions lines between the derived species components of branch xylem density (ρx ) and those 3.4.2 Branch xylem density 3.4.3 Leaf area: sapwood area ratio for mass per unit area (MA ), foliar [P] and foliar [K] for the same As detailed in Table 1, the derived taxonomic component Reasonably strong correlations were found for log10 (8LS ) species the average mass (S) for the associated genus. Red of ρx and was negatively correlatedseed with log [P], log [Ca], with log10 (MA ), log 10 10 10 [N] and log10 (LA ) (p ≤ 0.001) with (`A ) and positively associated withon log10 (S) fertility the relationship log10blue (8LS ) and log10 [P] also sig10 [K], log10 openlogcircles indicate species found low sitesbetween and the (p ≤ 0.001). A weaker but significant positive correlation nificant (p ≤ 0.01). The relevant biplots are shown in Fig. 5. openwascircles indicate species found on highThe fertility sites. Species also observed with log10 (M correlation slope for the taxonomic component MA ↔ 8LS relationA ) and a negative with log10 [Mg] (p ≤ 0.01). Of minor significance was a negship is 1/−1.27 = −0.79. Thus, as 8LS increases across found both with soillogfertility groups are indicated circles ativeon association species,with then Mclosed A declines proportionally less. That is to 10 (LA ) (p ≤ 0.05). Some of these illustrated in Fig. black 4 which shows rela- show say, species a higher 8LS also (seerelationships text for are details). The solidthelines the with SMA model fittend to carry a greater tionships between ρx and both [P] and [K] to be particularly weight of (generally larger) leaves per unit stem area with which did not depend on for soil fertility. compelling and, as is also the case MA and S, with no difthose leaves also tending to have higher foliar [N] and [P]. ference for species associated with low versus high fertility soils.

Biogeosciences, 9, 775–801, 2012

www.biogeosciences.net/9/775/2012/

Leaf nitrogen (mg g-1)

(a)

783

(b)

(d)

(c) Leaf area (m2)

Leaf phosphorus (mg g-1)

Leaf mass per unit area (g m-2)

˜ et al.: Tropical tree trait dimensions S. Patino

Leaf area:sapwood area ratio (m2 cm-2) Fig. 5. Standard Major Axis (SMA) regression lines between the derived species components of leaf area/sapwood area ratio (8LS ) and those for mass per unit area MA , foliar [N], foliar [P] and average leaf size for the same species.Red open circles indicate species found on low fertility sites and the blue open circles indicate species found on high fertility sites. Species found on both soil fertility groups are indicated with closed circles (see text for details). Solid lines show the SMA model fit which did not depend on soil fertility.

Fig. 5. Standard Major Axis (SMA) regressions lines between the Leaf nutrients and other structural traits 3.5 Common Component modelling derived3.4.4species components of leaf area/sapwood area Principal ratio (Φ LS ) (taxonomic components) Strongfor positive correlations (p ≤ 0.001) and those mass per unit areawere Malso foliar [N], foliar [P] and averA , observed for log10 (LA ) with log10 [N] and log10 [P] as well as between Results from the CPC modelling are shown in Table 2, with age leaf forS. the same both species.Red open circles indicate species log10size [Ca] and Interestingly, the slope and intercept the full model output, details of the rationale for eigenvecof these relationships are dependent on the soil fertility with tor inclusion and assessments of the overall model fit given found which on low fertility sites and the blue open circles indicate species a species is associated (Supplementary Information in the Supplementary Information Tables S3, S4 and S5 and found onsites. low fertility soils tend tofound have their The five eigenvectors selected found Table on S2B). highSpecies fertility Species on accompanying both soilcaptions. fertility a higher LA at any given foliar [N] and/or [P]. are listed in Table 2 in order of their importance, as derived groups are indicated with circles (see text forcharacteristic details).rootsSolid For the [Ca] ↔ S pairing the closed negative slope is also large from the (eigenvectors, λ). These results (−8.3), though in this case with no soil fertility effect decan be interpreted as in the case of an ordinary principal comlines show the SMA model which not depend on soil fertility. tected. Though not shown in Fig. 6,fit also of note isdid the posiponents analysis, the difference here being that the relative tive [C]↔ S relationship (p ≤ 0.001) with species with a low seed mass also tending to have a low foliar carbon content.

www.biogeosciences.net/9/775/2012/

weightings (λ) have been allowed to differ for species on high vs. low fertility soils. The first eigenvector, U1 , had somewhat higher λ for high vs. low fertility associated species (accounting for 0.24 and 0.27 of the dataset variance respectively) and with high positive coefficients for all three foliar cations and to a lesser Biogeosciences, 9, 775–801, 2012

imensions

˜ et al.: Tropical tree trait dimensions S. Patino

784

Table 2. Common principal component analysis of derived genetic effects for species associated with low and high fertility soils. Values in brackets represent standard errors for each component. Coefficients given in bold are either those whose absolute values are 0.50 or more, or 0.30 or more with a standard error of less than 0.1. MA = leaf mass per unit area; elemental concentrations are on a dry weight basis, LA = leaf area; 8LS = leaf area:sapwood area ratio, ρx = branch xylem density,  = diffusion limitation index (see Eq. 1), S = seed mass, Hmax = species maximum height. Variable log(MA ) [C] log[N] log[P] log[Ca] log[K] log[Mg] log(LA ) log(8LS ) ρx  log(S) Hmax Characteristic roots λlow,j λhigh,j

U1

U2

Component U3

U4

U5

−0.22(0.05) −0.35 (0.05) 0.15 (0.10) 0.25 (0.08) 0.42 (0.03) 0.48 (0.02) 0.49 (0.04) −0.01 (0.09) −0.01 (0.07) −0.14 (0.03) 0.10 (0.04) −0.23 (0.03) −0.10 (0.04)

−0.23 (0.06) 0.24 (0.07) 0.53 (0.04) 0.45 (0.05) −0.13 (0.08) −0.01 (0.09) −0.21 (0.09) 0.48 (0.05) 0.29 (0.06) −0.03 (0.05) 0.14 (0.05) 0.01 (0.06) 0.07 (0.06)

0.44 (0.07) 0.01 (0.07) −0.02 (0.09) 0.12 (0.09) 0.15 (0.09) 0.00 (0.08) 0.07 (0.06) 0.25 (0.13) −0.44 (0.11) −0.22 (0.10) 0.39 (0.09) 0.19 (0.10) 0.53 (0.10)

−0.22 (0.11) 0.06 (0.12) 0.22 (0.09) 0.31 (0.05) −0.31 (0.06) 0.16 (0.09) 0.06 (0.07) −0.35 (0.16) −0.53 (0.16) 0.12 (0.21) −0.10 (0.13) 0.48 (0.10) −0.13 (0.22)

0.35 (0.09) 0.34 (0.09) −0.03 (0.08) 0.08 (0.06) 0.00 (0.08) 0.05 (0.11) 0.19 (0.07) −0.16 (0.10) 0.18 (0.11) 0.26 (0.11) 0.60 (0.08) 0.59 (0.08) −0.47 (0.09)

1876 (259) 2341 (237)

1472 (203) 1641 (166)

641(89) 898 (91)

717 (99) 564 (57)

698 ( 96) 318 ( 32)

Table 3. Bivariate relationships for the derived environmental component of the observed plant traits.Values above the diagonal represent the slope of the relationship (y axis as columns labels, x axis as row labels). Values below the diagonal represent the correlation coefficient. Values significant at P < 0.05 are given in bold. NS = no slope estimated as the relationship was not significant. For units and symbols, see Table 1.

7 log[N] log[P] log[Ca] log[K] log[Mg] log(LA ) log(`A ) log(8LS ) ρx  S. Pati˜no et al.: Tropical tree trait dimensions log(MA ) −1.06 NS NS NS −3.49 NS 0.57 any −given 0.31 species) with the NS results −4.97 shown in Table 4. −1.32 This sed before. It is thus here [C] 0.63 − −3.38 NS −15.86 NS −4.22 NS NS NS 4.10 1.82 shows that 0.33 theseem total to variation in4.68 the 11 traits examined any given 3.06 species) with the NS results shown in T doesofnot have been recognised before. It is thusNS here log[N] −0.52 S.−0.30 − 2.69 NS 1.25 NS NS Pati˜noS. et Pati˜ al.: nthe Tropical o etfirst al.:PCA tree Tropical trait (ů tree dimensions trait dimensions could be explained by axis ) with ρ an imshows that 0.33 of the total variation in the 11 tra 1 x denoted0.48 as PFL .− −0.04 −0.09 1.74 1.53 NS NS NS NS −0.45 0.20 cted, all of whichlog[P] we beportant contributor and this also relating positively to foliar could be explained by the first PCA axis (ů 1 ) w log[Ca] −0.28 −0.54 0.28 0.50 − 0.88 0.27 NS NS NS −0.26 NS t (see Supplementary InOverall the five eigenvectors selected, all of which we beany given any species) given with species) the with results theshown results inshown Table not does not to have seem been toall have recognised been recognised before. Itbefore. isNS thus It here is thus here [C] anddoes M butseem negatively with foliar nutrients examined portant contributor and this also relating positi −0.01 −0.13 0.23 0.74 0.49 − NS NS NS −0.30 0.13 A ,lieve he total variance log[K] for both to be physiologically relevant (see Supplementary Inshows that shows 0.33 that of the 0.33 total of variation the total variation in the 11 in traits theex 1 and alsodenoted with . The second axis of the PCA on the plot eflog[Mg] −0.54 −0.72 0.28 0.04 0.50 0.05 − NS NS NS NS NS [C] and M , but negatively with all foliar nutrie as denoted .as PFL . for 0.68 of the total variance for both A PFLaccounted formation), could be could explained be explained by the first by PCA the first axis PCA (ů ) axis with (ů ρ log(LA ) 0.08correlation −0.06 −0.09 −0.03 0.11 0.21 0.07 − 0.85 NS NS NS 1 fects matrix (ů ) is also significant, accounting and also with . The second axis of the PCA o Overall the Overall five eigenvectors the eigenvectors selected, selected, all of which all of wewhich be- we be2 fivesoil low and high fertility species. es (normalised to ± 100) portant contributor and this also and this relating also positively relating p log(`A ) 0.07 −0.20 −0.16 −0.11 0.09negative 0.20 0.15 0.90 portant − contributor NS −0.79 −0.35 for 0.25 of the variance, with substantial weightings fects correlation matrix (ů ) is also significan lieve to be lieve physiologically to be physiologically relevant (see relevant Supplementary (see Supplementary InIn2 pplementary Information The 0.36 first three axes species scores to ±0.07 100)[C] 0.08 log(8LS ) −0.29 −0.21 (normalised 0.10 −A NS and [C] but negatively M , butNS negatively with foliar with nutrients all foliar nu ex A , and for MAformation), , −0.25 foliar [C] and against −0.07 (and to0.68 a−0.02 lesser extent [P])variance forM 0.25 of the variance, withall substantial negativ formation), accounted accounted for of forthe 0.68 total of foliar variance the total for both for both lack of any systematic are plotted each other in Supplementary Information ρx 0.08 0.27 −0.22 −0.64 −0.46 −0.82 −0.06 −0.25 and −0.31 0.17 −Theaxis NS also with and also . The with second . second of the axis PCA of the on the PC being balanced positive weightings for foliar [Mg] in parfor MA , foliar [C] and  (and to a lesser exte low0.27 and high low and fertility high soil fertility species. soil species. cores as expectedfor the Fig.by S1. shows the0.25 required any systematic 0.32 0.24This 0.49 0.31lack of −0.22 −0.14 fects −0.28 −0.09 −0.18 − )significant, correlation fects correlation matrix (ůmatrix also (ů acc 2 ) is weightings 2 is also ticular, but also with contributions from Φ and ρ . being balanced by positive forsignifi foliar LS x The first The three first axes three species axes scores species (normalised scores (normalised to ± 100) to ± 100) ciple components model. correlations between the species scores as expected for thefor 0.25 of for the 0.25 variance, of the variance, with substantial with substantial negative we but also with contributions from ΦLS ne an are plotted areagainst plotted each against other each other in Supplementary Information Information ns of these three trait dioutput from any good fitinofSupplementary a principle components model.for Mticular, , for foliar MA[C] , foliar and  [C](and andtoa(and lesser to extent a lesser fo A Fig. S1. Fig. This S1. shows This the shows required the required lack of any lack systematic of any systematic 8a also showing that it is Clearly a wide range of combinations of these three trait di-being balanced being balanced by positive by weightings positive weightings for foliar for [Mg f correlations correlations between between the plot species the scores species asscores expected as expected for that the itfor the 3.8 negative Relationship between effect PCAs and typically associated with [P], and mensions can for occur. But[C] with Fig. 8a also showing isticular, extent foliar coefficients foliar with coefficients for foliar [N] and [P] aswith well as with LA contributions and, from to a ΦLSfrom but ticular, also but also contributions and Φ ρ x L output from output anyspeaking) from goodany fit of good a principle fit of a principle components components model. with model. 3.8 Relationship between plot effect . soil/climate (generally only species typically associated and but or both PDJ smaller lesser extent, 8LS . Also notable are modestly negative coRW . still significant coefficients for MA and S. In Clearly aClearly wide range a wide of range combinations of combinations of these three of these traitthree di- trait disoil/climate that haveseems scores forefficients both PDJ onents of theterms three of major cations, carbon and high MA , fertility this firstsoils component forand MA RW and. foliar [Mg]. In terms of [N], [P] and mensionsmensions can occur.can But occur. with But Fig.with 8a also Fig.showing 8a also showing that it is that it is diagrammaticsimilar form. to This 7 shows major components threetomajor that first described byFigure Poorter andspeaking) dethe Jong (1999) Mtypically U2the , seems components of what isbetween considA ,of 3.8some Relationship 3.8 Relationship between plot effect plot PCA effe (generally (generally speaking) only species only typically species associated associated with reflect with The most significant relationships between the PCA site axis o be “shared”, CPCs and their(PDJ) overlap of traits in diagrammatic form.leaf This andespecially thus we dub it the Poorter-De Jong dimension, ered the classic economic spectrum (Reich et al., 1997; soil/climate soil/climate scores of Table 4, and previously soil climate most significant relationships between the and and . . Thelabel high fertility high soils fertility soils havecalculated that scores have forscores both for both PDJ PDJ RW RWthus illustrates thatthat many traits seem to and be “shared”, especially . , RW Wright et al., 2004). We this the Reich-Wright dir all three of PDJ PDJ characteristics of7the samethe are the shown in Fig. 9. First, the scores of Table 4, and previously calculated soi Figure Figure shows 7sites shows major components major components of the three of the major three major mension, , of tropical tree functional trait coordination. M which isůanas important factor forfirst all soil threePCA of RW , RW A9 the component, top panel ofU Fig. shows function ofdiagrammatic the PDJ ∩ RW ) and 1for CPCs and their and overlap theiraan ofoverlap traits in of traits in diagrammatic form.PDJ This form. This characteristics of the same sites are shown in Fig Theofsecond accounts additional 2 , CPCs most The significant most significant relationships relationships between the between [Mg] varying0.18 in opposite and .many Also occurring in be ( PDJ ∩Although of theTheseem axisdataset of illustrates Fyllas etillustrates al. (2009), the latter considered atostrong in-) and top panel of Fig. ů1 as function ofPCA the FW RW that that traits many seem traits to seem “shared”, be “shared”, especially especially and 0.19 of the variances for low and high fertilHmax would to have little9 shows influence ona eiscores of scores Table 4, of and Table previously 4, and previously calculated calculated soil and ese two traitity dimensions. same sign is [C], but with [P] and [Mg] varying in opposite . The strong tegratedM measure of soil fertility and denoted axis of Fyllas et al. (2009), the latter considere F species respectively, and is characterised by high positive ther or it emerges as the dominant term for U PDJ RW 3 Mis which important is an important factor forfactor all three for of all three A which A an PDJ ,of RW PDJ , characteristics RW characteristics of the same of the sites same are shown sites are in shown Fig. FF directions with respect to M for these two trait dimensions. relationship observed suggests a strong integrated response tegrated measure of soil fertility and denoted9. in A n the same direction relaand tropical and . Also . Also occurring in fertility, ( PDJinwith ∩( PDJ )∩and of )the and oftop thepanel top of Fig. panel 9 shows of Fig. ů 9 shows as a function ů as a of function the first of s FW FWoccurring RW RW 1 1 of Amazon forest trees to soil most nurelationship observed suggests a strong integra a high estimated standard Intersecting RW and FW and in the same direction relasame 2012 signsame is [C], sign butisfoliar with [C], but [P] with [Mg] [P] and varying [Mg]invarying opposite in opposite of axis of et Fyllas al. (2009), etforest al.the (2009), latterto the considered latter consid a stw Biogeosciences, 9, 775–801, www.biogeosciences.net/9/775/2012/ trients increasing, with [C]and andwith ρx adecreasing as of Fyllas Amazon tropical trees soil fertility, tive to and MA is ΦLS . Although high estimated standardaxis included LA in ( RW ∩ directions directions with respect with to respect M for to these M for two these trait dimensions. two trait dimensions. . Th tegrated measure tegrated of measure soil fertility of soil and fertility denoted and denote A A this plot of ů F increases. Interestingly, the Kendall’s τ for trients increasing, and with foliar [C] and ρ F x d error as part of FW , we have also included LA1 in ( RW ∩relationship ies in the opposite direcrelationship observed observed suggests suggests athe strong aintegrated strong Intersecting and in same in the direction same direction relarela- F increases. versus Intersecting is greater than forand any of theand original variInterestingly, Kendall’s τ forinttr RW and RW FW FW F of 0.63 ), this also showing that varies the opposite of Amazon of Amazon tropical forest tropical trees forest to soil trees fertility, to soil with fertilm FW tive to M tive to ΦLS M . Aet Although isal. ΦLS . Although with with estimated ain estimated standarddirecstandard ables examined by (2009), theaithigh highest ofhigh which versus A isFyllas RW cf. FW . F of 0.63 is greater than for any of the Variable

log(MA )

[C]

any given species) with the results shown in Table 4. T shows that 0.33 of the total variation in the 11 traits examin denoted as PFL . could be explained by the first PCA axis (ů 1 ) with ρx an Overall the five eigenvectors selected, all of which we beportant contributor and this also relating785 positively to fo ˜ et al.: Tropical S. Patino tree trait dimensions lieve to be physiologically relevant (see Supplementary In[C] and M , but negatively with all foliar nutrients examin A formation), accounted for 0.68 of the total variance for both and also with . The second axis of the PCA on the plot low and high fertility soil species. along with MA and, of opposite sign, 8LS . Also of note here fects correlation matrix (ů ) is also significant, account 2 The first three axes species scores (normalised to ± 100) high value for the coefficient of the diffusion is the relatively for 0.25 of the variance, with substantial negative weighti are plotted against each other in Supplementarylimitation Information index,  which is positively associated with both for MA , foliar [C] and  (and to a lesser extent foliar [ Fig. S1. This shows the required lack of anyHmax systematic and MA . Interestingly, for this component LA varies being balanced by positive weightings for foliar [Mg] in p correlations between the species scores as expected for the in the opposite direction to 8LS (albeit with a large stanticular, but also with contributions from ΦLS and ρx . S. Pati˜ no et al.:components Tropical tree trait suggesting dimensions that there is a tendency towards conoutput from any good fit of a principle model. dard error) Clearly a wide range of combinations of these three trait disiderably fewer but also significantly larger leaves in taller mensions can occur. But with 8a also it is any but given species) with the res does Fig. not seem to showing havestatured beenthat recognised before. is thus here species. There It also being a modest significant 3.8 Relationship between plot effect a (generally speaking) onlydenoted species as typically associated with shows that 0.33 of the PCAs total varia negative contribution of ρx to this dimension. We consider PFL . soil/climate could be explained by the first P . its own high fertility soils that have scores forthe both U3and , which on accounts for be0.08 and 0.10 of the variOverall five eigenvectors selected, all of which we PDJ RW portant contributor and this als Figure 7 shows the major components of theation threein major lieve to be physiologically relevant (see Supplementary the dataset respectively, toIn-contain several features [C] and M negatively with Pati˜nino diagrammatic et al.: Tropical tree trait dimensions A , butfor CPCs and their overlap ofS. traits form. This formation), accounted for 0.68 of the total variance for bothand Westoby similar to those described by Falster (2005) The most significant relationships between the PCA site a and also with . illustrates that many traitslow seem be fertility “shared”, andtohigh soilespecially species. climax tropical forest in Australia, and it is thus denoted asThe second ax scores of Table 4, and previously calculated soil fects correlation matrix (ů clim is 15 20 25 does 30 2 )res any given species) withand the not seem to have been recognised before. It is thus here The for firstallthree (normalised to ± 100) . , scores FW MA which is an important factor threeaxes of species PDJ RW -1 characteristics of the same sites are shown in Fig. 9. First, for 0.25 of the variance, with su shows that 0.33 of the total varia Foliar [N] (mg g )denoted are plotted against each The otherfourth in Supplementary Information axis is dominated by S and 8LS PFL ∩ . RW ) and of the component and FW . Also occurring in ( as topofpanel ofsystematic Fig. 9 shows ůcould a be function theand firstsoil P for foliar of [C] (and 1 asM A , explained Fig. S1. PDJ Thisfive shows the these required lack any with coefficients of different sign. Associated with the by the first P Overall the eigenvectors selected, all of which we besame sign is [C], but with [P] and [Mg] varying in opposite axis of Fyllas et al. (2009), the latter considered a strong being balanced by positive weig portant[P] contributor correlations between higher the species as expected for Inthe to these be physiologically relevant (see lower Supplementary S arescores also [Ca] but higher foliar and LA . and this also directions with respect to lieve MA for two trait dimensions. . Thewith stro tegrated measure of soil fertility and ticular, butAdenoted also contributio F [C] and M , but with negatively output from any good fit of a principle components model. formation), accountedWith for 0.68 of the totalfor variance for both and higher standard lower values their coefficients relationship observed suggests a strong integrated respo Intersecting RW and FW and in the same direction relaand also withterms. . The second ax Clearly a wide range soil of combinations threesign, trait are di- the M low and high fertility species. errors, also beingof ofthese different [N] A and of Amazon tropical forestfects trees to soil fertility, with(ůmost tive to MA is ΦLS . Although with a can high estimated standard correlation matrix S. Pati˜ n o et al.: Tropical tree trait dimensions 2 ) is mensions occur. But with Fig. 8a also showing that it is The first three axes As species scores (normalised to ± 100) mentioned intrients the Discussion, U forand 0.09 4 (accounting increasing, and with foliar [C] ρ decreasing x for 0.25 of the variance, with su 3.8 Relationship between (generally speaking) species typically associated with error as part of FW , we are have also included LAonly in ( ∩ RW plotted against each other in Supplementary Information and 0.07 of the population variance for low and high fertility τ forand this plot of F increases. Interestingly, forthe MKendall’s , foliar (and soil/climate A any given[C] species)with t does notin seem tospecies haverequired been recognised before. ItbeisRW thus here This shows the lack of any systematic and . highitS1. fertility soils have scores for both that varies thethat opposite direcrespectively) seems to dominated by the PDJ FW ), this also showing Fig. versus F of 0.63 is greater thanbalanced forpresence any of the original v being by positive weig shows that 0.33 of the tota correlations between the species scores as expected for the 7 shows the of the major seeded members ofthree the Leguminaceae imporas FW . largecomponents PFL . ofmajor tion relative to MA and ΦLS Figure fordenoted ables examined by Fyllasticular, et al.whose (2009), the highest of wh RW cf. but also with contributio be explained by the output from anyoverlap good fitof oftraits a principle components model. CPCs and their in phytogeography diagrammatic form. This tance in the of Amazon has with already Overall the five eigenvectors selected, allfor of foliar which we Comparison be-forestcould was 0.56 [P]. Fyllas et al. (20 The most significant relationshi portant contributor and th Clearly a wide range of combinations these three trait illustrates that many traits seem relevant to beofalso “shared”, especially lieve to be physiologically (see Supplementary Inbeen recognised by ter Steege et al. We therefore 3.6 Bivariate relationships: environmental components shows that thediů 2 (2006). contains significant weightings of le scores[C] of and Table previous M4, but negative mensions can occur. But with Fig. 8aall also it for isindividually, A ,and formation), accounted for 0.68 of theshowing total both denote this dimension as variance .that TS MA which is an important factor for three of ,that, level variables were all strongly correla PDJ RW characteristics of the. same 3.8 Relationship between and with Thesites sec (generally speaking) only species typically associated with low and highfertility fertility soil species. data from1.5 both low and high sites The lasttoeigenvector included inthe our analysis, U5Aalso , )differs 0.5 Considering 1.0 2.0 with mean annual precipitation (P viz.9positive . Also occurring in ( ∩ ) and of and top panel of Fig. shows ů1corre as a( FW PDJ RW soil/climate fects correlation matrix -1 high fertility gether, Foliar Table 3[P]lists correlations and SMA slopes for the enThe first three axes species scores (normalised to ± 100) and have [P] scores for both others intions having substantially greater importance (mg g ) same sign is soils PDJ ain RW with foliar [C].and M a negative correlation A and [C], that but from with and [Mg] varying opposite axis offorFyllas et al. (2009), thew 0.25 of the variance, w vironmental effects with this information provided in more are 7plotted against other in Supplementary Information Figure shows the for major of trait the three low fertility versus high fertility speciesnot (accounting for foliar [Mg]. It major is therefore surprising, asofissoil shown in directions with respect toeach MAcomponents for these two dimensions. tegrated measure fertility for M , foliar [C] and  detail (including confidence intervals) in the Supplementary Fig. S1. This shows the0.04 lack ofform. any systematic CPCs and their overlap of traits inrequired diagrammatic This 0.09 and of the population This second panel ofvariances Fig. 9, thatrespectively). ů 2 and PA A also show strong as relationship observed suggests being balanced by positiv Intersecting andtraits and in be the“shared”, same direction rela-for the The most significant relationshi FW Information (Table S2A). illustrates As forcorrelations Table the between SMA slopes rethe species scores as expected that1, RW many seem to especially ciation, but by withHexamination ofhaving Tabletropical 4 also suggesting t component is characterised MAAmazon oppomax and of forest trees but4,also with cont tive tooutput MAxisas Φthe Although with a high estimated standardmodel. scoresticular, of Table and previousl LS . any flect the relationship y ↔ x, with the column headers from good fit of principle for any given species, Φ and ρ also decline w site signs (in tocomponents ) ,and with both higher S and  also LS w FW MA which is an important factor foracontrast all three of PDJ RW trients increasing, with fol characteristics of theand same sites S. Pati˜ nFW o et S.,range al.: Pati˜ Tropical nof oassociated et al.:tree Tropical trait dimensions tree trait dimensions error as part we have also included L in ( three ∩; this and the y being the row labels. For theof traits, the Clearly astructural wide combinations trait diRW increasing precipitation and, somewhat counter intuitiv being withof a Athese lower H also being along max increases. Interestingly, thea . Also occurring in ( ∩ ) and of the and F panel top of Fig. 9 shows ů as FW PDJ RW 1 most significant relationshipsFW are all and appear mensions can occur. with Fig.with 8a showing that it is  increasing. ), thisisnegative also showing that itbevaries thealso opposite direcwith aBut less substantial but significant coefficient forF ρof Alsois greater tha x . 0.63 versus same[Ca], sign [C], speaking) but with [P] and [Mg]invarying in opposite axis of3.8 Fyllas etgiven al. (2009), thew Relationship betw tween ρx and log10 [P], log log to a lesser only species typically associated with any any species) given does not and, seem does to not have seem been to have recognised been recognised before. It before. is thus here Itfoliar is thus here 10 (generally 10 [K] oftoΦinfluence in characterising U are greater [C] asso5 cf. . tion relative to respect MA and for ables examined by Fyllas et al. directions with M for these two trait dimensions. RW FW LSA tegrated measure of soil fertility Finally, as in Fyllas et al. (2009) we show values soil/climate extent log10 (`A ). The slopes observed (−0.26soils tociated −0.41) shows thatshows 0.33 of that the and Although, . high denoted fertility that.have for both with the . higher MA and U5 for presents asdenoted asare,scores PDJ . RWwas PFL PFL foliar [P]. Compar relationship observed partial τ (denoted τ0.56 all traits ofsuggests interest Intersecting and for and in the Kendall’s same direction relap ) forcould however, much less than for the associated the taxRWslopes FW be explained could be ex b Figure 7 shows the major components of the three major some combinations as reported previously inwe the literOverall the Overall fivetrait eigenvectors the five eigenvectors selected, all selected, of which all we of which be-shows be3.6 Bivariate relationships: environmental also that the T ůforest 2,contains of Amazon tropical as components ů 1 and ů 2 as functions of portant Patrees and tive in to Table MA is1Φ(−0.37 with a highwell estimated standard F , T , contributor a onomic components as listed to −0.72). LS . Although portant con a CPCs and their overlap of traits in diagrammatic form. This lieve to believe physiologically to this be physiologically relevantmostly (see relevant Supplementary (see with Supplementary In-variables In-that, ature, component, related species found at τindividually level trients increasing, with fol in Table 5. ( Here∩the calculated value of and and asso p The most significant relat [C] and M [C] , but and nega M error as partformation), ofdatathat ,many we have also included L in A A FW A RW illustrates traits seem to be “shared”, especially Considering from both low and high fertility sites toformation), accounted accounted for 0.68 of for the 0.68 total of variance the total for variance both for both low fertility soils,ated doesprobability not seem togiving have been recognised bemean annual precipitatio increases. Interestingly, the anwith of the with effect of eK Findication 3.7 Principal component analysis of environmental efscores of Table 4, and pre and also and . also The w gether, Table 3 lists correlations and SMA slopes for the enlow and high low fertility and high soil fertility species. soil species. ), this also showing that it varies in the opposite direcfore. It is thus here denoted as PFL tions after with foliar [C] and Mthe an all three of , RWversus A PDJ. parameter 0.63correlation is greater than soil/environmental accounting F of 2 6 10 FW 14MA which is an important factor for fects characteristics of for thecorre same fects fects ma vironmental effects with this information provided inselected, more The first three The first axes three species axes scores species (normalised scores (normalised toTaking ± 100) to ±account 100) Overall the five eigenvectors all of which we befoliar [Mg]. It is therefore not -1 tion relative to M and Φ for cf. . ables examined by Fyllas et al. RW FW A LS fect of the other four. into to the poten and FW . Also occurring in ( PDJ ∩ RW ) and of the Foliar [Ca] (mg g ) top panel of Fig. 9 0.25 shows for 0.25 offor the varian ofů detail (including confidence intervals) ineach theSupplementary Supplementary are plotted are against each against other in other in Supplementary Information Information to be physiologically relevant (see Supplementary second panel of InFig. 9, that ů 2eta was 0.56 for foliar [P]. Compar Especially given the strong relationships ρplotted the effects of spatial autocorrelation (Fyllas same signbetween is [C],lieve but with [P] andconfounding [Mg] varying in opposite x and axis of Fyllas et al. (2009 for M , foliar for M [C] , fo a A A Information (Table S2A). As for Table 1, thethe SMA slopes re-systematic Fig. S1. Fig. This shows This shows required lack required ofof any lack ofrelationships any systematic ciation, butfor with examination os formation), accounted for 0.68 the total variance both 3.6 Bivariate components also shows that the ů 2being contains foliar cation environmental components (Fig.8) , itS1. was ofthe we only consider with pbalanced ≤ 0.01 or bet directionsrelationships: with respect toenvironmental M these two trait dimensions. tegrated measure of soil f A for 2009) being by balan po flect the relationship ylow ↔ x, with thefertility x as the column Fig. 6. Standard Major Axis (SMA) regression lines between decorrelations correlations between the between species the scores species asheaders scores expected as for expected the for thethat, any given species, both Φ and high soil species. level variables individually additional interest to see if coordinated structural/leaf bioAs for the (full) Kendall’s τforshown inticular, Fig. 9, but Table 5 sugge relationship observed sug ticular, also with but Intersecting and and in the same direction relaRW FW rived taxonomic components of foliar [N] and foliar [P] yand leafthe output and the being rowany labels. For the structural the (normalised output from good fit of good aaxes principle fitto ofbea traits, components principle components model. model. increasing precipitation and, s data both low and high fertility sites toThe first three species scores tothe ± 100) with mean annual precipitation chemical responses to the Considering environment existisfrom for Amazon for-any the superior predictors than individual variab of Amazon tropical forest tive to M Φ with a of high estimated standard Arelationships LS . Although mass per unit area (LA ) for the top two panels and between species most significant are allcombinations negative and appear beClearly a wide Clearly range a wide of range combinations of these three of these trait three ditrait diwith  increasing. gether, Table 3 lists correlations and SMA slopes for the enare plotted against each other in Supplementary Information with [C] and[K] M an est. We therefore undertook (S) a PCA analysis of the full plot the only exception being tions Ta . In thatfoliar case, [N], and A trients increasing, and w estimated foliar [Ca] associated average seed mass for the astween ρ and log [P], log [Ca], log [K] and, to a lesser error aseffects part10ofand ,S1. we have also included L in8a( also ∩ mensions can occur. can But occur. with Fig. But 8a with also Fig. that showing itany is [Mg]. that itItisisregressing x mensions FW Ashowing RW vironmental with this information provided in more 10 10 Fig. This shows the required lack ofpresent systematic foliar therefore not s effects correlation matrix (excluding H S both of all show relationships not when the max increases. Interestingl F 3.8 sociated genus (bottom panel). Open regressions circles indicate species found Finally, as inRelationship Fyllas al.p Fig. 6. Standard Major Axis (SMA) lines between derived taxonomic 3.8 etRelat extent log10 (`A ). The slopes observed (−0.26 tospecies −0.41) are,direc(generally (generally speaking) speaking) only only typically associated typically associated with with detail (including intervals) in the Supplementary also showing that itspecies varies inPCs the opposite correlations between the species scores as expected for the second panel of Fig. 9, that ů a FW ), thisconfidence which were considered to be environmentally invariant for effect as dependent variables. 2 versus of 0.63 is great F soil/climate fertility andfoliar the closed indicate species found Kendall’s partial τ (denoted τ soil/c however, less than for associated for taxomponentsonoflow foliar [N]sites and [P] circles and leaf mass permuch unit area (L )from for the1, top Information (Table S2A). As for Table theslopes SMA slopes reAthe andcomponents . examination high fertility high soils fertility that have soils scores that have for both for both ciation, but RW with of output any good fit scores of a.the principle model. PDJ PDJ RW .and tion relative to M and ΦLS for ables examined by Fyllas on high fertility sites. Species found on both soil fertility groups RW cf. to FW A well as ů and ů as function 1 2 onomic components as listed in Table 1 (−0.37 −0.72). the associated relationship ↔ x, with the x range asthe themajor column headers wo panels are anddesignated between species estimated foliarflect [Ca] mass for given species, both Figure y7average shows Figure the shows major components components of the three major the any three major Clearly a7 seed wide of combinations ofofthese three traitHere di- foliar was 0.56 for [P].ΦCL by a “+” (see text for details). For the top two panin Table 5. the calcula andspecies the3.6y being the row labels. For the structural traits, the S) for the associated genus. Open circles indicate found on low fertility sites increasing precipitation and, so CPCs and CPCs their overlap and their of overlap traits in of diagrammatic traits in diagrammatic form. This form. This mensions can occur. Figure 8a also shows that it is (generally Bivariate relationships: environmental components els, solid lines show the SMA fit for low fertility soil which also shows that the an ůmost 2 con ated probability giving inds 3.7 species Principal component analysis of environmental efThe most The significant most significant relationships are all negative and appear bewith  increasing. illustrates illustrates that many that traits many seem traits to be seem “shared”, to be “shared”, especially especially speaking) only species typically associated with high fertility nd the closed circles indicate species found on high fertility sites. Species found on are significantly different to the dashed lines for high fertility soil levelscores variables that, indivi soil/environmental fects of parameter Table scores 4,ofanT tween ρx and log10data [P],soils log [Ca], log [K] and, for to aboth lesser to10 10two Considering from both low and high fertility species.groups For the bottom panel the solid shows the text SMA model .the that have high scores with mean annual precip PDJ RW M whichMis which important is antop important factor for all factor three forofallsites three , of , oth soil fertility are designated bylines a ”+” (see for For the A details). Aan PDJand RW PDJ RW fect of other four. Taking Finally, as in Fyllas et al. characteristics characteris of the extent log (` ). The slopes observed (−0.26 to −0.41) are, A 10 Table whichshow did notthe depend on soil fertility. 3 strong listsFigure correlations andbetween SMA slopes forthethe en-of 7Also shows the three major tions with foliar and anels, solidfitlines SMA fit for low fertility soilgether, species which are significantly Especially given the relationships and confounding effects of[C] spatial and and . Also occurring . associated occurring inthe ( major inρ∩xcomponents (the )∩ and of the ) and of the Kendall’s partial τ (denoted τopM top panel of top Fig. panel 9 sh FW FW PDJfor RW PDJ RW however, much less than for the slopes taxvironmental effects with this information provided in more2009)foliar CPCs and their overlap of traits in diagrammatic form. This [Mg]. It is therefor foliar cation environmental components (Fig.8) , it was of we only consider relation same sign same is [C], sign but is with [C], [P] but and with [Mg] [P] and varying [Mg] in varying opposite in opposite ifferent to the dashed lines for high fertility soil species. For the bottom panel the solid well as ů 1 axis and of ů 2Fyllas asaxis functions et of al. Fyl( onomicdetail components as confidence listed in Table 1 (−0.37 to −0.72). (including intervals) in the Supplementary second panel of Fig. tha additional interest todirections see respect if coordinated structural/leaf bioAsTable for the5.(full) Kendall’s τ9,sho directions with with to respect MA for tothese MAtwo for these trait dimensions. two trait dimensions. in Here the calculat tegrated measure tegrated of ms nes shows the SMA model fit which did not depend on soil fertility. Information (Table S2A). As for Table 1, the SMA slopes reciation, but with examina chemical responses to the environment exist for Amazon forthe to berelationship superior predictors ated probability giving an indi relationshi observed 3.7 Principal component analysis of environmental efwww.biogeosciences.net/9/775/2012/ Biogeosciences, 9, 775–801, 2012 Intersecting Intersecting and and and in the and same in direction the same reladirection relaRW RW FW FW flect the relationship y↔ x, with the x asofthe column headersthe only for exception any givenparameter species, est. We therefore undertook PCA the fulla plot being Ta . bo soil/environmental of Amazon of tropical Amazon faI fects tiveytobeing M tive isthe ΦtoLS M . aA Although is ΦLSanalysis .For Although with a structural high with estimated high estimated standard standard A and the row labels. the traits, the increasing precipitation a effects correlation matrix (excluding Hmax and S both of all show relationships not pres fect of the other four. Taking trients increasing, trients incr an most significant relationships are all negative and appear beerror as part error of as part , we of have , also we have included also L included in ( L ∩ in ( ∩ FW FW A RW A RW with  increasing. which were considered to relationships be environmentally invariant for effect PCs asFeffects dependent variabla Especially the log strong the confounding ofF spatial increases. increase Interes x andto tweengiven ρx and [P], log [Ca], between log [K]ρand, a lesser 0.010

0.010

.1.0 0.1

Seed mass (g)

10

0.001

Leaf area (m2) Leaf area (m-2)

0.100

0.001

Leaf area (m2)

0.100

does not seem to have been recognised before. It is thus here

˜ et al.: Tropical tree trait dimensions S. Patino

786

Д PDJ

Д RW

[K] [P] [C] [Mg] [Ca]

[N] LA Φ ρ

Table 4. Summary of the Principal Components Analysis of the correlation matrix for the derived environmental/soil effects on observed structural and physiological traits. Coefficients given in bold are those whose values are 0.3 or more. MA = leaf mass per unit area; elemental concentrations are on a dry weight basis, LA = leaf area; 8LS = leaf area: sapwood area ratio, ρx = branch xylem density,  = diffusion limitation index (see Eq. 1). Variable

M

Component

S. Pati˜S. noPati˜ et al.: no Tropical et al.: Tropical tree treedimensions trait dimensions 7 7 A trait ů1 ů2 ati˜ aitndimensions oS.etPati˜ al.:nTropical o et al.: Tropical tree traittree dimensions trait dimensions 7 7 7 LS log(MA ) −0.196 −0.443 any given any species) given species) with the with results the results shownshown in Table in 4. Table This 4. This does not does seem notto seem havetobeen haverecognised been recognised before.before. It is thus It ishere thus here [C] −0.300 −0.412 shows shows that 0.33 that of 0.33 the total of the variation total variation in the 11 in the traits 11 examined traits examined any given species) with the any results given any shown species) given in species) with Table the 4. with results This the shown results in shown Table in 4. Table This 4. This ognised notdoes seem before. not to seem have It been is to thus have recognised here been recognised before. It before. is thus It here is thus here denoted denoted as XPFL as . PFL . 0.320 0.111 log[N] tree Tropical trait dimensions tree. trait Overall dimensions 7 the could be could explained bevariation explained by11 the by first PCA first axis PCA (ůaxis (ů ρx an imshows that 0.33 of the total variation shows in0.33 thethat 11 of the traits 0.33 total of examined the variation total in7the intraits the 11 examined traits examined 1 ) with 1 )ρwith x an imoted denoted as Overall the fivethe eigenvectors fivemax eigenvectors selected, selected, all ofshows which all ofthat which we bewe belog[P] 0.406 −0.276 PFLas PFL . portant portant contributor contributor and this and also this relating also relating positively positively to foliar to foliar could be explained by the first could PCA be could explained axis be (ů explained ) with by the ρ first an by imPCA the first axis PCA (ů ) axis with (ů ρ ) an with imρ an im1 x 1 1 x x lieve bebephysiologically to be physiologically relevant relevant (see (see In- In- log[CA ] 0.453 0.099 verall selected, the Overall five all ofeigenvectors the which fivetolieve eigenvectors we selected, selected, all of which all ofwe which be- Supplementary we be-Supplementary [C] and [C] M and , but M negatively , but negatively with all with foliar all nutrients foliar nutrients examined examined e trait dimensions S. Pati˜ n S. o et Pati˜ al.: n o Tropical et al.: Tropical tree trait tree dimensions trait dimensions 7 7 7 portant contributor and this portant also relating contributor portant positively contributor and this to and foliar also this relating also relating positively positively to foliar to foliar A A any given any species) given with species) the results with the shown results in shown Table 4. in Table This 4. This en o have recognised been recognised before. It before. is thus here It is thus here log[K] 0.392 −0.300 formation), formation), accounted accounted for(see 0.68 for of0.68 theInof total thevariance total for both for both eevant to lieve be(see physiologically toSupplementary be physiologically relevant Inrelevant (see Supplementary Supplementary In- variance and also and with also . with The . second The second axis of axis the of PCA the on PCA the on plot the efplot ef[C] and M , but negatively [C] with and all M [C] foliar , and but nutrients M negatively , but examined negatively with all foliar all nutrients foliar nutrients examined examined shows that shows 0.33 of that the 0.33 total of variation the total in variation the 11 traits in the examined 11 traits examined A A A log[M ] 0.245 0.416 low and low high and fertility high fertility soil species. soil species. g mation), of. the formation), total accounted variance accounted forfor 0.68 both for of the 0.68total of the variance total variance for both for both FL fects correlation fects correlation matrix matrix (ů is(ůalso )7is significant, also significant, accounting accounting and also with . The second and axis also of and with the also . PCA with The on second . the The plot axis second efof the axis PCA of the on the plot the efplot efno Tropical et al.: Tropical tree trait tree dimensions trait dimensions 27)PCA 2on could be explained could be explained by the first by PCA the first axis PCA (ů ) with axis (ů ρ an ) with imρ an im0.087 −0.009 log(L ) any given species) with the results shown any given in any Table species) given 4. species) This with the with results the shown results in shown Table in 4. Table This 4. This recognised does before. not does seem It is not thus to seem have here to been have recognised been recognised before. before. It is thus It here is thus here 1 x 1 x A and es. low high and fertility high soil fertility species. soil species. The first The three first axes three species axes species scores scores (normalised (normalised to ± 100) to ± 100) vectors five eigenvectors selected, all selected, of which all we of which be- we befor 0.25 for 0.25 the variance, of the variance, with substantial with substantial negative negative weightings weightings fects correlation matrix (ů fects is correlation also fects significant, correlation accounting (ů matrix )of also significant, isthe also significant, accounting accounting log(8 )(ůfoliar 0.025 0.271 portant contributor and this also and relating this also positively relating positively to tototal foliar 2 ) variation 2 2 )of shows 0.33 the total in shows thematrix 11 that shows traits 0.33 examined that ofis the 0.33 total variation variation in the 11 in traits the 11 examined traits examined FW LS are plotted are against against each other each in other Supplementary inofto Supplementary Information Information he scores first The (normalised three first axes three toaxes ± 100) scores species (normalised (normalised to ±portant 100) ±contributor 100) lly siologically relevant (see relevant Supplementary (see InIn-that denoted denoted asspecies asplotted .Supplementary . scores PFL PFL for M for , foliar M , [C] foliar and [C]  and (and  to (and a lesser to a lesser extent extent foliar foliar for 0.25 of the variance, with for substantial 0.25 for of the 0.25 negative variance, of the variance, weightings with substantial with substantial negative negative weightings weightings ρ −0.383 0.287 A [C] and M [C] ,lack but and negatively M ,given but negatively with all foliar nutrients all foliar examined nutrients examined xAbe could be explained by the first PCA axis could (ů be with explained ρx explained an by imthe first by the PCA first axis (ů 1axis ) with (ů ρ an imρx an[P]) im- [P]) AInformation A ot o seem have to been have recognised been recognised before. before. It iseigenvectors thus Itselected, here isrequired thus here any given any species) species) with the with results the shown results in shown Table in4.Table This 4. PCA This 1 )could 1 )x with Fig. S1. Fig. This S1. shows This shows the the required of lack any ofsystematic any plotted nors Supplementary areof against plotted each against other each in other Supplementary in Supplementary Information rounted 0.68 forthe 0.68 total variance the total for variance both for both selected, Overall all ofInformation which Overall five we the eigenvectors befive selected, all of which all of we which bewesystematic be 0.174 −0.340 being balanced being balanced by positive by positive weightings weightings for foliar for [Mg] foliar in [Mg] parfor M , foliar [C] and  for (and M to , for a foliar lesser M [C] , extent foliar and [C] foliar  (and and [P]) to  a (and lesser to a extent lesser foliar extent [P]) foliar [P]) Fig. 7. Euler diagram showing overlaps between the first three also with and also .scores The with second The axis second of the axis PCA of on the the PCA plot on efthe plot efA A A portant contributor and this also relating portant positively portant contributor to contributor foliar and this and also this relating also relating positively positively to foliarto foliarin parshows that shows 0.33 that of the 0.33 total of the variation total variation in the 11 intraits the 11 examined traits examined correlations between between theand species the scores as expected as. expected for the for the uired S1. lack This S1. of shows This any systematic shows required lack of lack any systematic ofspecies any systematic lrelevant species. soil (see lieve Supplementary tocorrelations lieve bethe physiologically to bethe physiologically In-required relevant relevant (see Supplementary (see Supplementary InIndertility as . Fig. . species. FL PFL ticular, ticular, but also but with also contributions with contributions from Φ from and ΦLS ρxexamined and . ρ . dimensions for the individual measured traits (where significant): being balanced by positive being weightings balanced being for balanced by foliar positive [Mg] by positive weightings in parweightings for foliar for [Mg] foliar in [Mg] parin parLS fects correlation fects correlation matrix (ů matrix ) is also (ů significant, ) is also significant, accounting accounting [C] and ,bebut negatively with allboth foliar [C] and nutrients [C] Mfirst and ,axis but examined M negatively negatively foliar nutrients foliar6.23 nutrients examined 2 for 2by could be could explained be explained by the first the PCA PCA (ůA1,axis )but with (ůρ1with )x with an all imρwith imA AEigenvalue output output from any from good fit good of aM fitwe principle offor avariance principle components components model. model. x an all elations es scores correlations between as expected between the accounted species for the the scores species scores expected as expected the for the 2.54 x pecies ee scores species (normalised scores (normalised to ± 100) toas ± 100) .68 of the formation), total variance formation), for both accounted for 0.68 for of 0.68 the total of the total variance for both five allaxes the eigenvectors five eigenvectors selected, selected, all of which allany of we which beblue; positive relationship with dimension, red; negative relationticular, but also with contributions ticular, but ticular, from also Φ but with and also contributions ρ with . contributions from Φ from and Φ ρ . and ρ . for 0.25 of for the 0.25 variance, of the with variance, substantial with substantial negative weightings negative weightings LS x LS LS x x and also with . The second axis of and the PCA also and with on also the . plot with The ef. second The axis second of the axis PCA of the on PCA the plot on the efplot efportant portant contributor contributor anddithis and also relating also relating positively positively to foliarto foliar Clearly aahigh wide asoil range wide of range combinations of combinations of these of three these trait three trait di-this principle ut output any components from good any fitand good of model. principle fit (see of aInformation principle components components model. model. other inst each in Supplementary other inClearly Supplementary Information ecies. low and low high fertility fertility species. soil species. siologically befrom physiologically relevant (see Supplementary Supplementary InInship withrelevant dimension, black; offor different sign depending on the diProportion of variance explained 0.33 0.25 M , foliar for M [C] , foliar and  [C] (and and to  a (and lesser to extent a lesser foliar extent [P]) foliar [P]) fects correlation matrix (ů ) is also fects significant, correlation fects accounting correlation matrix (ů matrix ) is (ů also ) significant, is also significant, accounting accounting Awith A [C] and [C] M and ,±but M negatively , but negatively with foliarall nutrients foliar nutrients examined examined 2100) 2 2 mensions mensions can occur. can occur. But with But Fig. 8aFig. also 8a showing also showing that it that is it is all with A± nations rly aClearly wide of these range alack wide three range combinations trait of dicombinations of these of three these trait three ditrait dihe shows required the(normalised required any lack of any systematic es scores The first three to first ± axes 100) three species axes scores (normalised scores (normalised toA to 100) on), ounted accounted for 0.68 for ofofThe 0.68 the total ofsystematic the variance total variance forspecies both both mension. Abbreviations are as in for Table 5,variance, with the three trait dibeing balanced being by balanced positive by weightings positive weightings for foliar [Mg] for foliar in par[Mg] in parfor 0.25 of the with substantial for 0.25 negative for of the 0.25 weightings variance, of the variance, with substantial substantial negative negative weightings weightings 3.8 Relationship 3.8 Relationship between plot effect effect PCAs PCAs and and and also and with also . with The . second The axis second of the axis PCA of the on PCA the between plot onwith the efplot ef- plot (generally (generally speaking) speaking) only species only species typically typically associated associated with with Fig. sions 8a can also occur. showing can But occur. with that But it Fig. is with 8a also Fig. showing 8a also showing that it is that it is tween species the scores species as scores expected as expected for the for the er inmensions Supplementary are plotted are against plotted Information against each other each in other Supplementary in Supplementary Information Information dertility high fertility soil species. soil species. mensions as defined in Sect. 3.5 ticular, but ticular, also with but also contributions with contributions from Φ from and ρ Φ . and ρ . for M , foliar [C] and  (and to a for lesser M extent for , foliar M foliar , [C] foliar and [P]) [C]  (and and  to (and a lesser to a extent lesser foliar extent [P]) foliar [P]) soil/climate soil/climate LS x LS x fects correlation fects correlation matrix (ů matrix ) is also (ů ) significant, is also significant, accounting accounting 3.8 Relationship between 3.8 Relationship plot 3.8 effect Relationship PCAs between between and plot effect plot effect PCAs PCAs and and A A cies erally (generally speaking) speaking) only species with only typically species typically associated associated with with tfirst yequired ofgood atypically principle fit of S1. aassociated principle components components model. model. Fig. lack of Fig. any This S1. systematic shows This the shows required the required lack of lack any of systematic any and systematic ee axes three species axes species scores (normalised scores (normalised to ± 100) to ± 100) and. 2 RW . 2 A high fertility high fertility soils that soils have that scores have scores for both for both PDJ PDJ RW being balanced by positive weightings being for foliar balanced being [Mg] balanced by in positive parby positive weightings weightings for foliar for [Mg] foliar in [Mg] parin par3.7 Principal component analysis of for 0.25 for of the 0.25 variance, of the variance, with substantial with substantial negative negative weightings weightings soil/climate soil/climate soil/climate combinations range of combinations these three trait didiecies scores correlations asof correlations between for the the species the species scores as scores expected expected for the themajor inst ted each against other each inexpected other Supplementary inof Supplementary Information Information and .between andPDJ and . as . of ores fertility high for both soils fertility that soils have that scores have for scores both forthe both Figure Figure 7these shows 7three shows thetrait major major components components of RW the three thefor major three PDJ RW PDJ RW ticular, but also from ticular, Φ ticular, and ρlesser with contributions with contributions from Φ from andΦρLS for with MAcomponents for ,contributions foliar M , foliar and [C]  (and and  to (and aalso to aalso extent lesser foliar extent [P]) foliar [P]) environmental effects LSbut x .but LS x . and ρx . A [C] occur. with But with 8a Fig. showing 8alack also that showing it isof that itsystematic isthree a principle output components from output any from model. good any fittheir good aany fit principle of aof components model. 1. shows This the shows the required of lack any systematic CPCs CPCs and their and overlap overlap of traits ofprinciple in traits diagrammatic in diagrammatic form.model. form. This This omponents gure 7Fig. Figure shows ofrequired 7also the the shows three major the major components major components of the the major three major being balanced being balanced by positive by positive weightings weightings for foliar for [Mg] foliar in [Mg] parin parThe most The significant most significant relationships relationships between between the PCA thesite PCA axis site axis 3.8 Relationship 3.8 Relationship between between plot effect plot PCAs effect and PCAs and aking) species only typically species associated typically associated with with mbinations ofoverlap these Clearly a species wide three range atrait wide diof range combinations of combinations of to these of“shared”, three trait three di- traitespecially ditween between theClearly species the scores as scores expected astraits expected for the for the illustrates illustrates that many that many traits seem traits seem be tothese be “shared”, especially Cs syions in and CPCs diagrammatic their and their form. of overlap traits This of in traits diagrammatic in diagrammatic form. form. This illustrates that many seem to beThis “shared”, especially Given the correlations between the environmental effects for ticular, but ticular, also but with also contributions with contributions from Φ from and Φ ρ . and ρ . scores scores of Table of 4, Table and 4, previously and previously calculated calculated soil and soil climate and climate The most significant relationships The most between The significant most the significant PCA relationships site relationships axis between between the PCA the site PCA axis site axis soil/climate soil/climate LS LS x x Fig. 8a mensions also showing mensions can occur. that can But occur. isseem But Fig. with 8a Fig. also 8a showing also showing that it of is,that it is from yth any fit good of athat principle of aM principle components components model. model. trates m to illustrates be that “shared”, many traits many especially seem traits to be be “shared”, and and .“shared”, . especially ve ilsgood scores that have for scores both for both M which is important factor for all three of PDJ PDJ RW RW Afit PDJ M which isitan important iswith anto important factor for factor allespecially for three all three , RW A which A an PDJ PDJ RW, RW ρ and several foliar nutrients (Table 3; Fig. 8), it was of x characteristics characteristics of the same of the sites same are sites shown are shown in Fig. in 9. Fig. First, 9. the First, the scores of Table 4, and previously scores of calculated scores Table of 4, Table soil and and previously 4, climate and previously calculated calculated soil and soil climate and climate 3.8 Relationship between plot 3.8 effect Relationship 3.8 PCAs Relationship and between between plot effect plot effect PCAs PCAs and and pecies typically (generally (generally associated speaking) speaking) with only species only species typically typically associated associated with range a for wide of range combinations of combinations of these of three these trait three ditraitof ajor ws the components components ofFW the three of major the three major and . factor occurring in (three ∩ ) ,andRW of) with the same PDJ RW or which M all ismajor which three an important of isand an important for factor all three for all of ,Also , diA PDJ RWAlso PDJ RW PDJ . . occurring Also occurring in ( in ( ∩ ∩ and ) of and the of the and additional interest to see if coordinated structural/leaf biotop panel top of panel Fig. of 9 shows Fig. 9 shows ů as a ů function as a function of the first of the soil first PCA soil PCA FW FW PDJ PDJ RW RW characteristics of the same characteristics sites are characteristics shown of in the Fig. same 9. of First, the sites same the are sites shown are in shown Fig. 9. in First, Fig. 9. the First, the soil/climate soil/climate soil/climate 1 1 occur. ns canBut occur. with But Fig. with 8afertility also Fig. 8a showing also showing that it[Mg] isscores that it for is PDJ of overlap traits inhigh ofboth diagrammatic traits in diagrammatic form. This form. This and . is[P] andPDJ and .direc. scores for fertility high soils that soils have that scores have for both both sign [C], with and varying in PDJ RW RW RW same sign same is sign [C], [C], with but [P] and [P] [Mg] and varying [Mg] varying opposite in opposite ( FW . only Also ∩FW occurring .only Also )is and occurring ofbut inthe (typically inbut ∩ ( associated )∩with and of the and ofopposite the and chemical responses the environment exist for Amazon axis of axis Fyllas etaxis al.ato(2009), et al. (2009), the latter the considered latter considered a strong a forstrong in- intop panel ofmost Fig. 9) shows ůsignificant as panel ainfunction top of panel Fig. of 9the of shows Fig. first 9soil ů1PCA shows as PCA aof function ůFyllas as function of the first of soil the first PCA soil PCA PDJ RW PDJ PDJ RW RW The The significant most relationships relationships between the between site the PCA site axis er diagram showing overlaps between the first three dimensions in terms 1top 1 3.8 Relationship 3.8 Relationship between between plot effect plot effect PCAs PCAs and and aking) lly speaking) species species typically associated with with many seem traits to Figure be seem “shared”, to be “shared”, especially especially reits components of the 7 Figure shows three 7 major the shows major the components major components of the three of the major three major tions with respect to M for these two trait dimensions. InA directions with respect with respect to M tofor M these for two these trait two dimensions. trait dimensions. and sign same [Mg] is [C], sign varying but isdirections [C], with in opposite but [P]with and [Mg] [P] and varying [Mg] in opposite in(2009), opposite est. We therefore undertook PCA analysis ofF .the full plotstrong The strong . The tegrated tegrated measure measure of soil of fertility soil fertility and denoted and denoted Avarying Aal. axis of Fyllas et axis the of latter Fyllas axis considered of et Fyllas al. (2009), a et strong al. the (2009), inlatter the considered latter considered a strong a instrong inF scores of Table scores 4, of and Table previously 4, and previously calculated calculated soil and climate soil and climate soil/climate soil/climate dual measured traits (where significant): blue; with raits insoils diagrammatic CPCs and CPCs their form. and overlap their This overlap ofFW traits of in traits diagrammatic in form.positive This form. This and and .thetrait . direction rtility ils that have that scores have for scores both for both tersecting and and in same relative torelationship PDJ PDJ RW RWdiagrammatic RW nt nctions important factor for factor all three for of all three of ,these , trait RW PDJ RW or these directions with two respect trait with dimensions. to respect M for to M for two these two dimensions. dimensions. effects correlation (excluding and Sresponse bothaxis of relationship observed observed suggests a strong aHPCA strong integrated integrated response . significant The strong . suggests The . The strong tegrated measure of soil fertility tegrated and tegrated measure denoted measure of soil fertility ofrelationship soil fertility denoted and denoted APDJ A max characteristics characteristics of the same ofespecially sites the same areThe shown sites in Fig. shown 9. First, inand Fig. the 9.matrix First, the Intersecting Intersecting and and and in and the in same the direction same direction relarelaFare F F The most significant relationships between most the The PCA most site significant axis relationships relationships between between the the site PCA axis site RW RW FW FW seem to be illustrates “shared”, illustrates that especially many that traits many seem traits to seem be “shared”, to be “shared”, especially re ws 7 the shows major the components major components of the three of the major three major M is 8 . Although with a high estimated standard error as d; negative relationship with dimension, black; of different sign dependA LS ) ∩ of Amazon of Amazon tropical trees forest totrees soil to fertility, soil fertility, with most with numost nuing so in occurring inthe ( same in ∩ ( RW and ofLS the )is and of thesame were considered toPCA beintegrated environmentally invariant for observed relationship strong relationship observed observed suggests suggests atropical strong aforest integrated strong response response top panel of top Fig. panel 9estimated shows of Fig. ůa91standard as shows acalculated function ůintegrated aof of function theresponse first of soil the PCA first soil PDJ PDJ RW RW tive to tive Mrelais FW M Φ . relationship Although Φ .the Although with arelawith high asuggests high estimated standard rsecting nd Intersecting and direction and and in the and same in direction direction relascores of Table 4, and previously scores scores soil Table and of 4, climate Table and previously 4, and previously calculated calculated soil and soil climate and climate 1 aswhich Ato A LS RW FW nd their overlap of[Mg] traits of inAtraits diagrammatic in diagrammatic form. This form. This part of ,in we have also included L inof (three ∩, Amazon ),of, this FW A all RW actor for all M three M isstandard which an important is an important factor for factor all three for of ,varying A PDJ RW PDJ RW PDJ RW mension. trients trients increasing, increasing, and with and foliar with [C] foliar and [C] ρFirst, and decreasing ρthe decreasing as as C], hoverlap [P] but and with [P] and varying [Mg] opposite in opposite of tropical forest trees of to soil Amazon fertility, tropical forest tropical with most forest nutothe soil trees fertility, to soil fertility, with most with numost any given with the results shown Table 4. This xnuxFirst, axis of Fyllas axis et of al. Fyllas (2009), etFW al. the (2009), latter considered the latter considered a species) strong inathe strong inwith to M ative high ismany to Φestimated M .which is Although Φto .be Although with apart high with estimated a FW high estimated standard standard characteristics ofalso the same sites are shown in characteristics Fig. 9.trees of First, the the same of sites same are sites shown are inshown Fig. 9.inin Fig. 9. the The most The significant relationships relationships between between PCA the site PCA axis site axis A LSseem A LS error as error part asof of ,Amazon we have ,the we also have included included Lmost L (significant in (characteristics ∩ RW ∩ FW A inrelative ARW many es that traits traits seem “shared”, to be “shared”, especially especially also showing that it varies in opposite direction increases. increases. Interestingly, Interestingly, the Kendall’s the Kendall’s τ for this τ for plot this of plot ů of ů orespect M for to these M two for these trait dimensions. two trait dimensions. trients increasing, and with trients foliar increasing, trients [C] and increasing, ρ and decreasing with and foliar with as [C] foliar and [C] ρ decreasing and ρ decreasing as as in (error ∩ )(Also and of .also occurring Also the occurring in included (tegrated inof (∩L ∩)ofscores and the and ofpreviously the and and that ofF 9.the total 11the traits examined F 1 1 .FThe strong strong measure tegrated of measure soil fertility of soil and fertility denoted and denoted A A FW x ofin top panel Fig. 9∩(shows ů∩of as aand function top panel ofxshows the top of first panel Fig. 90.33 of shows PCA Fig. ůThe shows axfunction ůvariation a function thethe first of soil first PCA soil PCA PDJ RW FW PDJ PDJ RW RW F soil scores Table 4, Table 4, and previously calculated calculated soil and soil climate and 1)of 1 as 1 asclimate also as part included of asfactor part LM ,Afor we in.factor have ,for we ∩ have included also L (that in FW Aof FW RW A in RW A it RWin ), this also this showing also that varies varies in varying the opposite the opposite direcdirecto and 8 for cf. . it FW FW RW FW LS nrP] ich important is an important three all of three of ,showing , and PDJ RW PDJ RW versus of 0.63 of is 0.63 greater isfirst greater than for than any ofa any the original increases. Interestingly, the increases. Kendall’s Interestingly, τsites for this plot the ofal. Kendall’s ůshown Kendall’s τFig. for this τconsidered plot for this ofaxis ůconsidered plot ů 1of and [Mg] same varying sign same in [C], sign opposite but is), [C], with but [P] with and [P] [Mg] varying [Mg] in opposite inthe opposite could be explained by the PCA (ů ) with ρstrong anvariimF F relationship relationship suggests aincreases. suggests strong integrated aInterestingly, strong integrated response response F axis Fobserved Fsame 1(2009), 1for of Fyllas etobserved al. (2009), the latter considered of Fyllas axis aofversus strong et Fyllas inetthe al. (2009), the latter the latter strong athe inin- varix original 1of characteristics characteristics of ofaxis the same are sites shown are in Fig. 9. in First, 9. the First, the and inFW the and same inisall the direction same reladirection relaRW FWand t ),varies this also in ),the this showing opposite also showing that direcvaries that in varies the opposite in the opposite direcdirecW FW tion relative tion relative toRW M toand M Φ)of and for Φ for cf. cf. .than .versus examined ables examined by Fyllas by Fyllas etoriginal al. (2009), et al. (2009), theThe highest the highest oftowhich of which versus of 0.63 is greater versus any of 0.63 the isof greater original 0.63 is greater varifor than any of for the any ofdenoted the original varivarifor these directions two trait directions with dimensions. respect with to respect for to M these for two these trait two dimensions. trait dimensions. RW RW FW FW A LS LS Amazon of tropical Amazon forest tropical trees forest to tegrated soil trees fertility, to with fertility, most numost nuso Also in (estimated in (∩itestimated ∩)AitM and of the and of the portant contributor and this also relating foliar F Fand The strong strong strong tegrated measure of soil fertility denoted tegrated measure measure of soilwith fertility of soil fertility and and denoted A A A top panel top of panel Fig. 9for ofshows Fig. 9ůofshows aables function ůsoil athan function of the first of the soil first PCA soil PCA FW PDJ RW ough Although with aoccurring high with aPDJ high standard standard F F .positively F . The 1 Fas 1 .as LS .occurring rrelative tion cf. to relative M and . to M Φ for Φ for cf. cf. . . was 0.56 was for 0.56 foliar for [P]. foliar Comparison [P]. Comparison with Fyllas with Fyllas et al. (2009) et al. ables examined by Fyllas et ables al. (2009), examined ables the examined by highest Fyllas by of et Fyllas which al. (2009), et al. the (2009), highest the of highest which of which RW FW RW RW FW FW A A LS LS trients increasing, trients increasing, and with and foliar with [C] foliar and ρ [C] decreasing and ρ decreasing as as C], gn but is [C], with but [P] with [P] [Mg] and varying [Mg] varying in opposite in opposite [C] and M negatively with all foliar nutrients examined observed suggests strong relationship integrated relationship observed response observed suggests suggests aainstrong a integrated strong integrated response response(2009) xthe xconsidered axis Fyllas axis ofet Fyllas al.a(2009), et al. (2009), the latter considered latter a strong strong inA , but and inwe the Intersecting same Intersecting direction relaand and in∩the andsame in theof direction same direction relarelaW RW FW FW e have have included also L included in (and LRW ∩ in (relationship FW ,also A RW A RW 3.6 Bivariate 3.6 Bivariate relationships: relationships: environmental environmental components components 3.6 Bivariate relationships: environmental components also shows also shows that the that ů the contains ů contains significant significant weightings weightings leafwas 0.56 for foliar [P]. Comparison was 0.56 was for with 0.56 foliar Fyllas for [P]. foliar et Comparison al. (2009) [P]. Comparison with Fyllas with et Fyllas al. (2009) et al. (2009) 2 2 increases. increases. Interestingly, Interestingly, the Kendall’s the Kendall’s τ for this plot τ for of this ů plot of ů ns respect with to respect M for to M these for two these trait two dimensions. trait dimensions. of Amazon tropical forest trees to soil of fertility, Amazon of with Amazon tropical most tropical forest nutrees forest to trees soil fertility, to soil fertility, with most with numost nuandfertility alsodenoted with of the PCA on the of plot ef-of leaf1 second 1axis . The strong strong measure measure of soil fertility of soil and and. denoted h with a high tive Ato estimated M tive toΦM standard . isAlthough ΦLS . Although withF a high withtegrated estimated aFhigh tegrated estimated standard standard FThe F . The A isA LSA goenvironmental that showing it varies that initcomponents the varies opposite in the direcopposite direcBivariate 3.6 Bivariate relationships: relationships: environmental environmental components components level variables level variables that, individually, that, individually, were all were strongly all strongly correlated correlated also shows that the ů contains also significant shows also that shows weightings the ů that contains the of ů leafcontains significant significant weightings weightings of leafof leafversus versus of 0.63 is of greater 0.63 is than greater for any than of for the any original of the varioriginal vari2 2 2 trients increasing, and with foliar [C] trients and ρ increasing, trients decreasing increasing, and as with and foliar with [C] foliar and [C] ρ and decreasing ρ decreasing as as fects correlation matrix (ů ) is also significant, accounting Frelationship F relationship suggests suggests ax stronga integrated strong integrated response x 2 response x cting andRW andas and in the and same in the direction relarelaRW also FW ave included error error part LFW of as ( part of ,∩ wesame have , direction we also have included also included L in (high Lobserved in ∩sites ( observed ∩ toA in FW RW FW A A RW RW Considering Considering data from data both from low both and low high and fertility fertility tosites with mean with mean annual annual precipitation precipitation (P ) viz. (P ) positive viz. positive correlacorrelalevel variables that, individually, level variables were level all variables that, strongly individually, that, correlated individually, were all were strongly all strongly correlated correlated Φ M for and Φ cf. for . cf. . A A ables examined ables examined by Fyllas by et al. Fyllas (2009), et al. the (2009), highest the of highest which of which increases. Interestingly, the Kendall’s increases. τ for this increases. Interestingly, plot of ů Interestingly, the Kendall’s the Kendall’s τ for this τ plot for this of ů plot of ů RW FW RW FW LS A LS of Amazon of Amazon tropical tropical forest trees forest to trees soil fertility, to soil fertility, with most with numost nufor 0.25 of the variance, with substantial negative weightings F F F 1 1 1 M . isAlthough ΦLS . Although with a high with estimated a high estimated standard standard LSA gether, Table 3and Table lists 3correlations lists correlations and SMA and slopes SMA slopes for the for enthe enow sidering Considering high data fertility from data both sites from low toboth low high and high fertility sites tosites toat itand varies inConsidering the ),gether, this opposite also ), this showing direcalso showing that itfertility varies that itand in varies the opposite in the opposite direcdirecdata from both low0.56 high fertility sites totions tions foliar with [C] foliar and [C] M and and M a negative and a negative correlation correlation with with FW FW with mean annual precipitation with mean (P with ) annual viz. mean positive precipitation annual correlaprecipitation (P ) viz. (P positive ) viz. positive correlacorrelaA A was for was foliar 0.56 [P]. for foliar Comparison [P]. Comparison with Fyllas with et al. Fyllas (2009) et al. (2009) A A A versus of 0.63 is greater than for any versus of the versus of original 0.63 of is varigreater 0.63 is than greater for than any for of the any original of the original varivaritrients increasing, trients increasing, and with and foliar with [C] foliar and [C] ρ and decreasing ρ decreasing as as for M , foliar [C] and  (and to a lesser extent foliar [P]) F F A F x x of ,SMA we3FW have , we also included also included LAcomponents incorrelations ( Lslopes inwith ∩(foliar ∩information FW RW RW vironmental effects effects with this information this provided provided intions more in the more er, spart and Table gether, Table slopes correlations 3.have for lists the correlations enSMA and SMA for slopes the enfor the engether, Table 3and and SMA slopes for the foliar [Mg]. foliar [Mg]. It is therefore It is therefore not surprising, not surprising, as is shown as is shown in the in the eships: relationships: environmental environmental components with [C] and M tions and with a negative foliar with [C] correlation foliar and [C] M and with and M a negative and a negative correlation correlation with with also shows also that shows the ů that contains the ů significant contains significant weightings weightings of leafof leaffor tion cf.lists relative tion relative tovironmental M tolists M ΦAtions forA Φ for cf. cf. . . A A A ables examined by Fyllas et al. (2009), ables examined highest ables examined of by which Fyllas by et Fyllas al. (2009), et al. (2009), the highest the highest of which of which 2 2 increases. increases. Interestingly, Interestingly, the Kendall’s the Kendall’s τ for this τ plot for this of ů plot of ů RW FW RW RW FW FW A and LSand LS being balanced by positive weightings for foliar [Mg] in parF F 1 1 (including detail (including confidence intervals) intervals) inthat, the in Supplementary thethat, Supplementary nmental vironmental provided with effects in with more information this information provided provided more in more onformation this showing alsoeffects showing that itdetail varies thatthis itin varies the opposite inconfidence the opposite direcdirecsecond second panel of panel Fig. of 9, Fig. that 9, ů that and ů P and also P show also strong show strong assoassoenvironmental effects with this information provided in more foliar [Mg]. Itinisversus therefore foliar not surprising, [Mg]. foliar It [Mg]. is as therefore is shown It is therefore not in the surprising, not surprising, as is shown as is in shown the in the 2 2 A A level variables level variables individually, individually, were all strongly were all correlated strongly correlated was 0.56 for foliar [P]. Comparison with was 0.56 Fyllas was for et 0.56 foliar al. (2009) for [P]. foliar Comparison [P]. Comparison with Fyllas with et Fyllas al. (2009) et al. (2009) versus of 0.63 of is greater 0.63 is than greater for than any for of the any original of the original varivariticular, but also with contributions from 8 and ρ . F F x LS Information Information (Table (Table S2A). S2A). As for As Table for 1, Table SMA 1,ůannual the slopes SMA slopes rereilM ervals) (including detail (including thedetail confidence Supplementary confidence intervals) in the Supplementary in Supplementary both ata from low both and high low and fertility high sites fertility tosites to-the ciation, with but with examination Table of 4Table also 4suggesting suggesting that that second panel of Fig. 9,mean that second and panel P second ofA panel Fig. show 9, of strong that Fig. ůbut 9, assoand that ůAthe also and P show strong show assostrong asso(including confidence intervals) in the Supplementary ps: 3.6 Bivariate 3.6 components Bivariate relationships: relationships: environmental components components with mean with annual precipitation precipitation (P )shows viz. (P positive )ciation, viz. correlapositive 2examined A also 2 2examination A also also shows that the ůthe contains also weightings also that shows the of that ůleafůhighest contains significant significant weightings weightings ofalso leafof leafative and to in M Φ for ΦLS for cf. . intervals) . environmental A ables examined bysignificant Fyllas by et Fyllas al. (2009), et al. (2009), the highest the of which ofof which 2ables 2Pcontains 2correlaRW RW FWcf. FW Aenvironmental A LSand flect the flect relationship the relationship ythe ↔ x, y with ↔ x, the with x as the the x column as the column headers headers or rmation Table Information 1, (Table theSMA SMA S2A). (Table slopes As S2A). for reTable As for 1, SMA 1, the slopes SMA reslopes reelations lists correlations and and slopes SMA for slopes the enforTable the enfor any for given any given species, species, both Φ both and Φ ρ and also ρ decline also decline with ciation, but with examination ciation, of Table but ciation, with 4 also but examination suggesting with examination of that Table of 4 also Table suggesting 4 also suggesting that that Information (Table S2A). As for Table 1, the SMA slopes reLS LSall w strongly w tions tions foliar with [C] foliar and M[C] and and aM negative and acorrelation negative correlation with with level variables that, individually, were level all variables level variables correlated that, individually, that, individually, were alleffect were strongly correlated 3.8 Relationship between plot PCAs andcorrelated with A[P]. Astrongly was 0.56 was for 0.56 foliar for foliar Comparison [P]. Comparison with Fyllas with etFyllas al. (2009) et al. (2009) and the and y being the y the being row the labels. row labels. For the For structural the structural traits, traits, the the the the flect relationship x as the the relationship column y ↔ x, headers with y ↔ the x, with x as the column x as the headers column headers fects this with information this information provided provided in more in more increasing increasing precipitation precipitation and, somewhat and, somewhat counter counter intuitively, intuitively, hth low and Considering high fertility Considering data sites from data toboth from low both and low high and fertility high fertility sites tosites tofor any given species, both for Φ any and for given any ρ species, also given decline species, both Φ with both and Φ ρ and also ρ decline also with decline with flect the relationship y ↔ x, with the x as the column headers foliar [Mg]. foliar It is [Mg]. therefore It is therefore not surprising, not surprising, as is shown as in is shown the in the ivariate relationships: relationships: environmental environmental components components LS LS LS w with mean annual precipitation with )w viz. positive withsoil/climate annual mean correlaannual precipitation precipitation (PwAof ) viz. (PApositive ) viz. positive correlacorrelaalso shows alsothat shows the that ů (P the ů 2 mean contains significant significant weightings weightings of leafleafA 2 contains most most significant relationships relationships are all are negative all negative and appear and appear bebe.the yand being thethe structural ythe being row the labels. row the For labels. the structural For the structural traits, the traits, the nce g For confidence intervals) intervals) in the Supplementary in the with  with increasing.  increasing. ions and gether, SMA slopes Table gether, for 3being Table lists the correlations en3Supplementary lists correlations and SMA and slopes SMA for slopes the enfor the enincreasing precipitation and, increasing somewhat increasing precipitation counter precipitation intuitively, and, somewhat and, somewhat counter counter intuitively, intuitively, second panel second of Fig. panel 9, that of Fig. ů 9, and that P ů also and show P also strong show assostrong assoand thetraits, ysignificant the row labels. For the structural traits, the tions with foliar and M and a that, negative tions tions correlation foliar foliar with and M andMacorrelated negative and a negative correlation correlation with with 2individually, Aindividually, 2with Awith level[C] variables level variables that, were all[C] were strongly all[C] strongly correlated A Aand A tween tween ρhigh and ρwith log and [P], log log [P], [Ca], log [Ca], [K] log and, [K] to and, ain lesser toTable lesser t.reAs significant all most negative significant relationships and appear relationships are beall negative are allinformation negative and appear and beappear bexin xhigh able S2A). for Table As 1, for the Table SMA 1, slopes the SMA reslopes reis information vironmental provided vironmental effects more effects this with this information provided provided inwith more more 10 10 10 10 10 10 with  increasing. with examination increasing. with a increasing. ata ering from data both from low both and low and fertility fertility sites tosites tociation, ciation, with examination but of 4The of also Table suggesting 4is also suggesting that that most significant relationships are allbut negative and appear bemost significant relationships between theisPCA foliar [Mg]. Itlog is therefore not surprising, foliar as [Mg]. foliar is shown It)[Mg]. therefore inApositive It the is therefore not surprising, not surprising, as is shown as in shown thesiteinaxis the with mean with annual mean precipitation annual precipitation (P viz. (P ) viz. positive correlacorrelaA Finally,Finally, as in as Fyllas in Fyllas et al. et (2009) al. (2009) we show we show valuesvalues for for extent log (` log ).10 The (` ).slopes The slopes observed observed (−0.26 (−0.26 toto −0.41) to −0.41) are, are, en Ca], tween log and [K] log ρx and and, [P], log to aextent [P], lesser [Ca], log [Ca], [K] log and, [K] to alog and, lesser to a lesser Alog A nship x,ρwith y3↔ the x, with as the the column x as the headers column headers x xtween intervals) detail in the (including detail Supplementary (including confidence confidence intervals) intervals) in the Supplementary in the Supplementary 10 10 10 10 10 10 10 10 Table lists correlations lists correlations and SMA and slopes SMA for slopes the enfor the enfor any given for any species, given both species, Φ both and Φ ρ also and decline ρ also with decline with second panel of Fig. 9, that ů and P second also show panel second strong of panel Fig. asso9, of that Fig. ů 9, and that P ů and also P show also strong show assostrong assoscores of Table 4, and previously calculated soil and climate ρlog and log [P], log [Ca], [K] and, a lesser LSA[C] w LSM wa negative tions with tions foliar with foliar and MAand and a Anegative and correlation correlation 2[C] 2 A 2 withA x 10 10 10 Kendall’s Kendall’s partial τ we (denoted τ (denoted τpshow ) forτpfor all ) for traits allfor of traits interest of interest as as Finally, asfor in Fyllas et al. Finally, (2009) Finally, as we in show Fyllas as in values etFyllas al. partial for (2009) et al. (2009) show we values values however, however, much much than less for than the associated the associated slopes slopes for the for taxthe taxbserved log extent (` (−0.26 log ).the The to (` slopes −0.41) ). (Table The observed are, slopes observed (−0.26 to (−0.26 −0.41) to are, −0.41) are, gfects labels. the row For labels. structural For the structural traits, the traits, the A A snt for Table Information 1, the Information SMA slopes S2A). (Table re-less As S2A). formore Table As for 1, Table the SMA 1,[Mg]. the slopes SMA reslopes 10 10information ental with effects this with this information provided provided in in more increasing increasing precipitation precipitation and, somewhat and, somewhat counter intuitively, counter intuitively, ciation, but with examination of Table ciation, 4not also ciation, but suggesting with but examination with that examination ofin Table ofin 4are Table also suggesting 4 alsoinsuggesting that9. First, that extent log slopes observed (−0.26 to −0.41) are, characteristics of the same sites Fig. foliar foliar It[Mg]. is therefore It isretherefore surprising, not surprising, as is shown as is shown the theshown 10 (`A ). TheKendall’s well as well ů and ů ů and as ů functions as functions of of , , T , , P , T and , P Q and Q partial τ (denoted Kendall’s τ ) for Kendall’s partial all traits τ partial (denoted of interest τ (denoted τ as ) for all τ ) traits for all of traits interest of interest as as 1 1 2 2 F T F a T a a a a a onomic onomic components components as listed as in listed Table in 1 Table (−0.37 1 (−0.37 to −0.72). to −0.72). p p p eships ever, associated however, much less slopes much than for less for the than the taxassociated for the associated slopes for slopes the taxfor the taxtincluding relationships are all negative are all and negative appear and beappear bethe xflect ashowever, the theflect column relationship the relationship ySupplementary ↔than x, with yfor ↔with the x, with x asincreasing. the the xcolumn as the column headers headers gwith confidence confidence intervals) intervals) inheaders theless in the Supplementary with slopes increasing. any given species, both Φ and forρůwany also for given species, given with species, both ΦLS both and ρassoand also ρdecline decline withfirst with second panel second ofpanel Fig. 9,of that Fig. 9, and that Pdecline ůany and also Pshow also strong show assostrong much for the associated for the top of Fig. 9Here shows ůΦ1calculated as function of of the soilassociLStaxLS wa w also 2the A 2panel AHere in Table in Table 5. 5. the calculated the value value of τ and τ associand well as ů and ů as functions well as of well ů and as ů as and functions ů as functions of of , , T , P and Q , , T , , P , and T , Q P and Q p p 1 2 1 F T 2 1 a a 2 a F T F a T a a a a a nmic Table components onomic 1 (−0.37 components as to listed −0.72). as in listed Table in 1 (−0.37 Table 1 to (−0.37 −0.72). to −0.72). og [P], [Ca], log log [Ca], [K] and, log to [K] a and, lesser to a lesser els. For and the structural the and y being the traits, y the being row the the labels. row For labels. the For structural the structural traits, the traits, the able ation (TableAs S2A). for Table Ascomponents for10 1,Table the SMA 1, as theslopes SMAreslopes re10S2A). 1010 increasing precipitation and, somewhat counter increasing precipitation intuitively, precipitation and, somewhat and, somewhat counter counter intuitively, intuitively, ciation, ciation, but to with but examination withincreasing examination ofated Table of4of Table also suggesting 4 giving also suggesting that that onomic in Table 1 (−0.37 −0.72). axis Fyllas et al. (2009), the latter a strong probability ated probability giving an indication indication ofconsidered the of effect the effect of each of each in listed Table 5.analysis the calculated in Table invalue 5.(2009) Table Here 5.τPCA the and Here calculated associthe calculated value ofvalue τpfor and ofan τassociassoci3.7 Principal 3.7 component component analysis of of environmental efefFinally, as Finally, inenvironmental Fyllas as inet Fyllas al. etofal. we (2009) show we values show forvalues p p and opes ).relationship The observed observed to (−0.26 −0.41) to are, −0.41) ps are most significant most and appear significant relationships berelationships areare, all negative areHere all and appear and beappear beenship yallslopes ↔negative x, with y(−0.26 ↔ the x, with x as the thePrincipal x column as the column headers headers with  increasing. with  increasing. with  increasing. for negative any for given any species, given species, both ΦLS both and Φ ρ and also ρ decline also decline with with LS w w soil/environmental soil/environmental parameter parameter after accounting after accounting for the for efthe efated probability giving an ated indication probability ated of probability the giving effect an giving of indication each an indication of the effect of the of effect each of each fects fects nalysis 3.7 oftween Principal environmental component component efanalysis of environmental oftraits, environmental efefKendall’s Kendall’s partial τto(denoted partial τlesser ) for of all interest traits ofasinterest as p ) for allτptraits for less the than associated for theρtween associated slopes for slopes the taxfor tax[Ca], log and tolog ρxanalysis aFor and lesser [P], log log [P], [Ca], log log [Ca], log and, and, a lesser toτ precipitation a(denoted g10 yPrincipal the being row the labels. row For labels. the structural the structural traits, the the xand, 10 [K] 10 10 10the 10 10 [K] 10 [K] increasing increasing precipitation and, somewhat and,offect somewhat counter counter intuitively, intuitively, fect the of other the four. other Taking four. Taking into account into account to the to potential the potential soil/environmental parameter soil/environmental after soil/environmental accounting parameter for the parameter efafter accounting after accounting for the for efthe effects fects well as(−0.26 ůbewell and(−0.26 asů−0.41) functions ů 2 (2009) as functions of Finally, ,Finally, T , PaFyllas and , for Tin P and(2009) Finally, as in Fyllas et al. we asavalues Fyllas etQaal. (2009) we show wevalues show for values for Biogeosciences, 9, 2012 www.biogeosciences.net/9/775/2012/ 1 2ů 1asand F , show Tof F, in Tas a ,Qet a a al. nents inasTable listed 1in (−0.37 Table to 1log −0.72). toThe −0.72). sgnificant observed extent (−0.26 log extent to −0.41) ).(−0.37 The (` are, slopes ).775–801, observed slopes observed to to −0.41) are, are, tsted relationships relationships are all negative are all negative and appear and beappear A A 10 (` 10 with between increasing. with  increasing. Especially Especially given the given strong the strong relationships relationships between ρ and ρother the and the confounding confounding effects effects ofassocispatial of autocorrelation autocorrelation (Fyllas et al., fect of the other four. Taking fect of into fect account the four. to other the Taking potential four. into Taking account into account to thespatial potential totraits the potential xpthe xof in Table in 5. Table Here 5. the Here calculated the calculated value of value τ and of associτ and Kendall’s partial τ (denoted τ ) for Kendall’s all traits Kendall’s of partial interest τ partial (denoted as τ (denoted τ ) for τ all ) for all of traits interest of interest as(Fyllas as et al., p p p p the associated however, slopes however, much for less the much taxthan less for than the associated for the associated slopes for slopes the taxfor the taxρog and [P], log log [P], [Ca], log log [Ca], [K] log and, [K] to and, a lesser to a lesser x 10 10 10 10 10 10 foliar foliar cation cation environmental environmental components components (Fig.8) (Fig.8) , it was , it of was of 2009) 2009) we only we consider only consider relationships relationships with p with ≤ 0.01 p ≤ or 0.01 better. or ecially ationships Especially given between the given strong ρ the and relationships strong the relationships between between ρ and the ρ and the confounding effects of spatial confounding autocorrelation confounding effects (Fyllas of effects spatial et of al., autocorrelation spatial autocorrelation (Fyllas et (Fyllas al., et al., ated probability ated probability giving an giving indication an indication of the effect of the of effect each of each x x x well as ůin1 Table and ů12 (−0.37 as−0.72). functions as, well ůeta1 , and ůand ů12 and asQafunctions ů 2values as functions of , al. T Pasa (2009) ,, P and Qaa and Qabetter. ldog ent analysis of analysis environmental of environmental efefFinally, as to in −0.72). Fyllas asof inwell Fyllas (2009) al. we show we show for values for Fet T F , of T, T Fa Ta, T a, P in10 Table 1The (−0.37 onomic components toobserved −0.72). components as(−0.26 as in Table (−0.37 to ).component The (`Aslopes ).onomic observed slopes (−0.26 tolisted −0.41) tolisted −0.41) are, are,Finally, additional additional interest interest to see to if see coordinated if coordinated structural/leaf structural/leaf biobioAs for As the for (full) the Kendall’s (full) Kendall’s τ shown τ shown in Fig. in 9, Fig. Table 9, 5 Table suggests 5 suggests mponents rr,less cation foliar environmental cation (Fig.8) environmental , it was components of components (Fig.8) , (Fig.8) it was , of it was of 2009) we only consider relationships 2009) we 2009) only with we consider p ≤ only 0.01 consider relationships or better. relationships with p ≤ with 0.01 p or ≤ better. 0.01 or better. soil/environmental soil/environmental parameter parameter after accounting after accounting for the effor the efin Table 5. Here the calculated value in Table of in τ 5. Table and Here associ5. Here calculated the calculated value of value τ and of τ associand associKendall’s Kendall’s partial τpartial (denoted τ (denoted τp ) pfor τall for alloftraits interest of interest as as p p p ) traits much thanless forthan the associated for the associated slopes for slopes the taxfor the tax-

S

H

Ø

Д

thecontributions required lack anymatrix also The axissignificant, of the PCAaccounting on the plot with from Φof and ρx. .systematic LS with fects and correlation (ůsecond ) is also

ef2 being balanced by positive weightings for foliar [Mg] in par00) he species scores as expected for the fects correlation matrix (ů ) is also significant, accounting 2 negative ± 100) for 0.25 of the variance, with substantial ticular, but alsoweightings with contributions from ΦLS and ρx . on fit of a for principle components model. for 0.25 of the variance, with substantial negative weightings ip between plot effect PCAs and ˜ S. Pati no et al.: Tropical tree trait dimensions 787 M , foliar [C] and  (and to a lesser extent foliar [P]) rmation A between plot effect PCAs and tic etip frait combinations of three di- (and tofor Mthese [C]trait andweightings a lesser extentinfoliar A , foliar dimensions dimensions 7 7 beingfor balanced by positive foliar [Mg] par- [P]) etematic he ut with ticular, Fig. 8aKendall’s also showing that it is Table 5. partial correlation coefficient, τweightings contribution (plotin effect estimate) of each foliar property (conbeing balanced by positive for foliar [Mg] parP , for the environmental also with contributions from Φ and ρx . estimated as detailed in Maghsoodloo and Laszlo for the trolling forbut LStheir the effects of the other environmental predictors) with significance el. 3.8 (p Relationship plot ateffect PCAs and nly species typically associated with ticular, but also with contributions from Φ andthe ρxbetween .results LS Pallos (1981). Bold values indicate a very strong correlation < 0.001) and italics indicate significant correlations p