Alternative height strategies among 45 dicot rain forest species from tropical Queensland, Australia


Daniel S. Falster (e-mail


  • 1Potential height, which spans at least an order of magnitude across species, is considered an important indicator of light capture strategy. Still, it remains unclear how potential height is coordinated with other traits that influence height growth rate, stem persistence and performance in low light. We proposed that contrasting correlations between potential height and other plant attributes would be observed for sets of species selected to span two hypothetical axes of light availability within mature forest and time since disturbance.
  • 2We selected 45 perennial rain forest species in Australia's wet tropics to span gradients of light availability and successional status and measured potential height together with traits influencing light capture and regeneration strategy on mature individuals. The traits included leaf mass per area, leaf nitrogen, wood density, stem mass per length, branch mass fraction and seed mass.
  • 3Potential height was significantly correlated with numerous traits among species selected to span each of the two gradients. Height was positively correlated with leaf mass area−1, leaf nitrogen and seed mass and negatively correlated with leaf area ratio at the branch tip along both light and successional gradients. Height was positively correlated with wood density along the successional axis, with the opposite relationship along the light gradient.
  • 4Trait relationships differed in either slope or intercept between the two gradients, reflecting different strategic trade-offs. At a given height, shorter species in the successional gradient were characterized by lower leaf mass area−1, lighter wood, smaller seeds, lower leaf nitrogen and lower leaf area ratio at the branch tip than similar sized species along the light gradient.
  • 5The results of this study support the idea of two distinct, trait-mediated axes of coexistence among short and tall plant species within vegetation. In several cases, trait relationships were weak or non-significant when species groupings were merged, indicating the importance of separating out the two sets for comparative studies.


Plant height reflects strategy for securing carbon profit via light capture (Grime et al. 1988; Weiher et al. 1999; Westoby et al. 2002). As taller plants shade shorter plants, competition favours additional expenditure on stems, opening an evolutionary arms race for light (Givnish 1982; Iwasa et al. 1985; Falster & Westoby 2003). However, not all plants are equally tall: in tropical vegetation; for instance, species potential (or maximum) height ranges from 1 to over 50 m (Foster & Janson 1985). Consequently, much of the emphasis in recent work has been on understanding the trade-offs associated with increased height that allow shorter species to persist in vegetation (e.g. Thomas 1996; Thomas & Bazzaz 1999; Turner 2001; Kohyama et al. 2003; Poorter et al. 2003).

Overall, there is growing consensus that species spread out along (at least) two axes summarizing light capture strategy (Fig. 1, Loehle 2000; Pacala & Deutschman 1995; Westoby 1998; Thomas & Bazzaz 1999; Sterck et al. 2001; Turner 2001; Poorter et al. 2003). The first corresponds to a vertical light gradient, comparing species at a point in time. Taller species capture a disproportionate share of available light (Hirose & Werger 1995; Kohyama 1993) but are restricted either in leaf area ratio or by poor sapling survival and performance in low light conditions (Iwasa et al. 1985; Givnish 1988; Aiba & Kohyama 1997; Thomas & Bazzaz 1999; Kohyama et al. 2003), opening an opportunity for shorter species to succeed.

Figure 1.

Schematic illustrating a simplified model for vegetation dynamics and our design for sampling species. Species were selected to span two ecological gradients: (i) a light gradient, spanning the full range of canopy positions found in mature rain forest (representing alternative light environments); and (ii) a successional gradient, spanning the full range of successional stages following disturbance. Species in the successional set are defined as those requiring full irradiance at maturity, and are assumed to be absent from late successional, closed forest where species in the light set are found.

The second axis known to be important for coexistence among height strategies extends through successional time (Fig. 1, Huston & Smith 1987; Shugart 1984; Pacala & Rees 1998). Following disturbance (removal of above-ground biomass), early successional species gain access to light pre-emptively via rapid height growth and superior colonization of vacant space. The requirements for rapid height growth bring with them increased risks of herbivory (Coley 1988), pathogen infection (Augspurger & Kelly 1984), mortality (Loehle 1988; Dalling et al. 1998; Davies 2001) and decreased shade-tolerance (King 1994; Kitajima 1994; Davies 1998). Due to decreased longevity and poor performance in low-light environments, early successional species are prevented from monopolizing time spent at the top of the canopy.

Species coexistence along the vertical axis is facilitated via trade-offs in realized productivity between high- vs. low-light environments. Along the successional axis, coexistence arises via trade-offs between the pace of height gain vs. longevity and shade tolerance (Huston & Smith 1987; Pacala et al. 1996; Pacala & Rees 1998). In each case, the trade-off may be manifested strategically via variation in several other key traits. For the light gradient at a point in time, relevant traits are expected to be coordinated with potential height (Thomas 1996). Species replacement during succession, however, also corresponds to an overall increase in potential height (Fig. 1, Gómez-Popma & Vászquez-Yanes 1981; Navas et al. 2003) leading to an alternative prediction for trait correlations with height. In principle, the nature of these correlations (their slopes and intercepts on bivariate plots) could differ depending on which gradient was being considered, reflecting the different strategic trade-offs associated with each axis.

The primary aim of the current study was to compare trait relationships with potential height among sets of species selected to span successional and vertical light gradients (Fig. 1). Several authors have reported correlations with one or another trait among species spanning one of the gradients (e.g. Thomas & Bazzaz 1999; Kohyama et al. 2003), but to our knowledge the two gradients have not yet been explicitly compared. Our working hypothesis was that trait relationships (if present) would differ between sets of species spanning each gradient, with tighter (higher r2) relationships observed among species spanning one of the gradients than across the entire species complement.

trait descriptions

Several leaf, wood, architectural and reproductive traits were chosen for study (summarized below). These traits are informative about height strategy because of their influence on height growth, on longevity or on growth at low light. Our list is by no means exhaustive; it was limited by resources and by the design imperative to quantify traits across significant numbers of species in the field. Summaries of each trait and expected relationships with potential height are provided in Table 1.

Table 1.  Summary of traits and hypotheses in the current study. Signs indicate the expected relationship between traits and potential height within sets of species spanning successional and vertical light gradients (see Fig. 1 and text). Key-words summarizing the reason for the correlation are given, with details given in the text
TraitDescriptionHypothesized correlation with potential height
(a) Successional gradient(b) Vertical light gradient
LMALeaf mass area−1 (mg mm−2)+ ve: height growth rate(1) + ve: maximize light interception
(2) − ve: resource conservation
Nmass, areaNitrogen leaf mass−1 or area−1 (mg N mg−1 or mm−2)− ve: height growth rate+ ve: light level
WDWood density (mg mm−3)+ ve: height growth rate; longevity– ve: hydraulic conductance; vertical growth
Stem extension rate (mm mm−1 day−1), comprised of: – ve: height growth rate– ve: structural reinforcement
SMPL: stem mass length−1 (mg mm−1) + ve: vertical growth
LMPL: leaf mass length−1 (mg mm−1)  
LMF: leaf mass fraction (mg mg−1)  
LAR: leaf area ratio (mm2 mg−1)  
LNF: leaf nitrogen fraction (mg N/mg stem)  
BMFBranch mass fraction (mg branch/mg shoot)+ ve: height growth rate– ve: light interception
SMSeed mass (mg)+ ve: colonization/shade tolerance+ ve: height allometry
Other traits
TwXSATerminal twig cross-sectional area (mm2)  
ANDry mass gain per leaf nitrogen (mg mg N−1)  

Leaf mass per area (LMA; mg dry mass mm−2)

LMA is one of several intercorrelated leaf traits, representing a fast–slow continuum in leaf economics across species (Wright et al. 2004). Low LMA is associated with short leaf life span, high leaf nitrogen, high photosynthetic capacity, short nutrient residence times and high relative growth rates (reviews by Westoby et al. 2002; Reich et al. 2003). Low LMA species are capable of rapid height growth (Coley 1988; Reich et al. 1992), but as a result may encounter increased mortality. A positive correlation with height is therefore expected in the successional set. Low LMA also improves use of low light, through its effect on leaf area ratio of the plant and hence on light capture per unit biomass (Givnish 1988). Recent reviews (Walters & Reich 1999; Reich et al. 2003), however, suggest the opposite: that LMA is higher among shade-tolerant species, contributing to a resource-conservation strategy favoured when carbon budgets are marginal (King 1994; Kitajima 1994). Thus there are two possible predictions on the relationship between LMA and height along the light gradient.

Wood density (mg mm−3)

The amount of dry matter invested per volume of stem varies considerably across species (e.g. from 0.26 to 1.29 for 243 tropical tree species, Ter Steege & Hammond 2001). Low density facilitates rapid volumetric growth (Enquist et al. 1999; Suzuki 1999) and is associated with high hydraulic conductance (Hacke et al. 2001), but results in decreased structural stability (Niklas 1994), increased risk of pathogen infection (Augspurger & Kelly 1984), cavitation risk (Hacke et al. 2001) and decreased shade tolerance (Lawton 1984; Loehle 1988; Osunkoya 1996). Density is commonly adopted as an indicator of successional status (Lawton 1984; Ter Steege & Hammond 2001), so we hypothesize a positive correlation with height in the successional group. Existing data indicate the opposite pattern for the light gradient (Thomas 1996; Kohyama et al. 2003) due to the need for increased vertical growth and hydraulic conductance among taller species.

Extension costs

The biomass cost per length of stem seems fundamental to a species’ height strategy. Yet despite significant variation among species (Poorter & Werger 1999), this trait has received little attention to date. Three measures of the process of stem extension bear consideration. First is the amount of dry mass required to achieve a unit of stem extension. Second is the rate at which leaf mass (or area) can be deployed in conjunction with a unit of stem growth. Third is the manner in which one and two combine to influence the rate of stem extension. Early successional species are hypothesized to economise on stem biomass (Schippers & Olff 2000), thereby facilitating rapid growth. Similarly, plants higher in the canopy may need stronger reinforcement to withstand increased exposure to wind (Osada et al. 2002), suggesting a negative relationship between extension rate and height in both sets (Table 1).

For comparison between species, extension cost is most usefully quantified for a common length of stem. In the current study, extension costs are quantified at two scales for mature plants: at the branch tip and for the terminal metre of stem. Measurements at the branch tip are essential for describing the effect of LMA, wood density and twig cross-sectional area on stem extension, without additional influence from branching and stem thickening. Branching increases the biomass cost per unit extension, decreasing height growth rate (Kohyama 1987; Kohyama & Hotta 1990). Thus we hypothesize a positive correlation between branching and potential height along the successional gradient. Lateral spread can also reduce self-shading, thus improving carbon gain under low light and increasing shade tolerance (Horn 1971; Kohyama & Hotta 1990; Pacala et al. 1996), suggesting a negative relationship with height along the light gradient.

Seed mass (mg)

Species mean seed mass summarizes much variation in dispersal and establishment success (reviews by Leishman et al. 2000; Westoby et al. 2002). Early successional species tend to have small seeds (Foster & Janson 1985; Osunkoya 1996), thereby emphasizing seed output and colonizing ability. Late successional species have larger seeds, emphasizing survival in low light (Foster & Janson 1985; Leishman et al. 2000). Recent work has also demonstrated a tight positive correlation with height across large numbers of species (Moles et al. 2004). Short early successional species are therefore hypothesized to have smaller seeds than equivalent-height late successional species (Foster & Janson 1985), but with a positive relationship between seed mass and height predicted for each set.


sites and species

The study was carried out in tropical rain forest vegetation of north-eastern Australia (Tracey 1982; Webb & Tracey 1994). Tropical rain forest covers a region along the coast from 21 to 15° S and contains within it a diversity of structural types (Webb & Tracey 1994). We restricted our sampling to regions of complex mesophyll vine forest found at Cape Tribulation (16°06′ S, 145°27′ E, 25 m a.s.l.) and on the Atherton tablelands (17°07′ S, 145°39′ E, 800 m a.s.l.). Rainfall is high throughout the region (3500 mm year−1 Cape Tribulation, c. 2000 mm year−1 Atherton) supporting a moderate to high level of foliage cover (leaf area index: 4–5 m2 m−2). Local endemism is high, although there is considerable overlap in species composition among localities (Osunkoya 1996). Preference was given to sampling species found widely throughout the region.

Tropical rain forest species are thought to differentiate out along both light and successional gradients and were sampled accordingly. Successional species were defined as those requiring a high-light environment at maturity (large gap or top of the canopy). For simplicity, we also assume species replacement at the top of the canopy corresponds to an increase in potential height, such that height is coordinated with successional status (Fig. 1). Nineteen species, ranging from early to late successional status, were selected using published sources (Hopkins & Graham 1987; Osunkoya 1996; Hyland et al. 1999). Species selected to represent the light gradient establish and mature in closed forest, where successional species are assumed absent (Fig. 1). For inclusion in this set we required that a species was known to establish and persist at low light levels, as indicated either by published sources (Hyland et al. 1999; Osunkoya 1996) or by scientists familiar with the vegetation (A Graham, J Wells, CSIRO Atherton; R Jensen). Twenty-six species, experiencing a range of light levels at maturity, from high (canopy trees) to low light (understorey shrubs), were included.

It is important to note that there is a natural intersection between the two gradients being considered (Fig. 1). Tall, late successional species cannot be distinguished from the tallest species in the light gradient. Thus there were degrees of clarity in how species were allocated between the two sets, ranging from very clear for shorter species in both sets, to not really clear for tall long-lived canopy dominants in mature forest (nine species in total). Despite the uncertainty, we thought it best to divide the nine species between the two groups, in preference to a posteriori classification or double counting. Traits of the tallest species in each group are therefore expected to converge, partly reflecting the somewhat arbitrary partitioning of these species between groups, but also reflecting common aspects of the biology. Similar trait correlations with height, including the intersection of trait values, were observed if these species were excluded from analysis.

species traits

Height and stem diameter were recorded on a large number of individuals spanning a range of heights to enable us to estimate potential height. Measurements of key structural traits were then made on three healthy, mature (> 60% potential height) individuals per species, located in light environments consistent with their classification into light or successional sets (Fig. 1). Successional species were located in gaps, clearings and road edges and light gradient species in patches of dense, mature forest. For each plant, a single metre of stem measured back from the tip at the tallest point was removed for measurement of all structural traits. Consequently stem traits reported refer to values expressed in the terminal metre of stem on individuals close to the asymptotic height of the species. The Australian Canopy Crane Research Facility (ACCRF) at Cape Tribulation gave access to the canopies of taller species. All data were collected in November–December 2002.

Potential height

The potential height of species (hpot) was determined using a plot of height (H) vs. stem diameter (D) measured at 10% of height (Aiba & Kohyama 1996; Thomas 1996; Ishii et al. 2000). An asymptotic function of the form:

image(eqn 1 )

where hpot, a and b are constants, was fitted to observed data for each species using non-linear regression. Parameter values of hpot, a and b that minimized the residual sum of squares were selected using the Levenberg-Marquardt estimation with tolerance of 10−6 in SPSS ver. 11.0. Calculated in this way, hpot represents the average top height realized by individuals of each species, not the overall maximum height observed.

Height-diameter data from several data sources were merged for analysis. Data points collected during the current study (551), in conjunction with operations at the ACCRF (327), or during surveys of 17 permanent plots by CSIRO Tropical Forest Research Centre (794) were included in the total of 1572. Data collected by CSIRO and ACCRF were mostly for large trees, with d.b.h. (c. 1.3 m height) rather than diameter at 10% of height recorded for each individual.

Despite significant scatter for some species, there were indications of an asymptotic relationship in most cases. Several obvious outliers in the ACCRF and CSIRO data sets were not included during parameter estimation on the premise that these were individuals with a history of significant stem damage (e.g. from wind), resulting in unusually large diameters for a given height. This was verified for trees at the canopy crane site by visual inspection. In total 39 points were excluded from analysis. Undoubtedly some damaged individuals remain, leading to some underestimation of potential height and increased scatter (Ishii et al. 2000). We did not perceive either the inclusion of individuals with a history of minor damage or the use of d.b.h. as significant limitations to analysis, as the effect will have been to reduce rather than increase interspecific spread in potential height.

Leaf traits

Leaf size (mm2), leaf mass area−1 (LMA; mg mm2) and leaf nitrogen concentration (Nmass,%) were measured on the first five fully expanded leaves at the tip of each individual. Leaf size was calculated as the one-sided leaf area (flat bed scanner) and LMA as the leaf dry mass (oven-dried for 48 hours at 65 °C) divided by leaf size. Leaves from all individuals per species were pooled and finely ground for nitrogen analysis. Total nitrogen concentration (%) was measured using complete combustion gas chromatography by Waite Analytical Services, Adelaide. Narea (mg N mm−2) was calculated as Nmass × LMA. For species with compound leaves and distinct mobile leaflets (Argyrodendron peralatum, Castanospermum australe, Cardwellia sublimis, Gillbeea adenopetala, Melicope elleryana, Polyscias australiana), leaf traits were calculated on the leaflet, with the rachis considered to be functionally equivalent to a branch. For all other species the petiole was included in measurements of all leaf traits.

Wood density (dry mass/fresh volume, mg mm−3)

Wood density was calculated using 40–60 mm stem segments taken 250 mm and 1000 mm back along a branch from the branch tip. Fresh samples were refrigerated before processing. After removing bark material, the volume of each wood sample was determined using Archimedes’ principle (Hacke et al. 2000). Samples were submerged in a water-filled container on a balance. The weight change (mg) recorded during submersion corresponds to the mass of water displaced, which can be converted to a volume using the formula: displacement weight (mg)/0.998 (mg mm−3), where 0.998 mg mm−3 is the density of water at 20 °C. Samples were then dried for 4 days at 60 °C before weighing.

Seed mass (mg seed−1)

Mean oven-dried seed mass (including seed coat but excluding seed accessories) was estimated for all species with available field material. In total, collections for 17 of 47 species were made. Data for an additional 12 species were drawn from published (Osunkoya et al. 1994; Grubb et al. 1998) and unpublished (P. Juniper, CSIRO Atherton) sources.

Stem extension

To compare across species the efficiency with which height growth is achieved, several elements warrant consideration (Table 1). First, there is the amount of stem dry mass required to achieve a unit of stem extension (stem mass per length = SMPL, mg mm−1). Secondly, there is the rate at which leaf mass (or area) area can be deployed in conjunction with a unit of stem growth. This can be quantified as leaf mass (or area) per stem length (LMPL: mg mm−1). LMPL + SMPL gives the total mass per unit extension. Finally, there is the manner in which an allocation profile described by SMPL and LMPL influences potential extension rate, a function both of biomass expenditure per length and of the expected revenue arising from the deployment of leaf area. To investigate this we set up the following model that partitions factors capable of influencing shoot extension rate. The model behaves as if all revenue and expenditure streams operate only within the terminal branch segment. If there is net export or import of photosynthate from the terminal segment, that will appear as lower or higher dry mass gain per unit leaf area of the terminal segment. Given a particular set of traits, potential stem extension rate for a branch segment of given length (SER: mm mm−1 day−1) can be partitioned as follows.

image(eqn 2 )

All terms except dry mass gain per leaf nitrogen (AN) can be compared across species from our data. The product of the last two gives leaf mass/shoot dry mass, which is equivalent to leaf mass fraction normalized per unit length (LMF, mg mg−1 mm mm−1). Similarly, the product of LMF with LMA−1 gives leaf area ratio (LAR, mm2 mg−1 mm mm−1). Multiplying by Narea gives leaf nitrogen per mass in the terminal metre or leaf nitrogen fraction (LNF: mg N mg−1 mm mm−1). One final multiplication by AN gives the estimated height growth per unit time (mm mm−1 day−1). Variation in each of these components is investigated. The questions of interest are: How variable is each component? When multiplied together do any of the components cancel each other out due to cross-correlation? Finally, how is the observed variation co-ordinated with variation in potential height?

SER is expressed per unit length because LMPL and SMPL vary with the length of the branch segment sampled, due to differences in leaf retention, stem thickening and branching. Consequently comparisons among species need to be made for similar-sized stem segments. We chose two scales for comparison. First, we quantified the most terminal branch segment sampled (up to 250 mm length), back to the first lateral branch or rachis. This provides an estimate of the allocation profile realized at the stem tip only. Structural traits including terminal twig cross-sectional area, LMA and wood density are expected to have important influences on SER. Secondly, we quantified the terminal metre of main stem. In addition to traits influencing variation at the tip, leaf retention and degree of branching will influence differences at 1 m length.

These costs measured at the tip or in the terminal metre do not, of course, represent the complete costs of an increment of height. Vascular increments down the full length of the stem and root increments could only be quantified through complete allometric analysis for each species.

Leaf retention and branching index

Leaf retention was calculated as the distance from the stem tip back to the oldest leaf along the main stem. The degree of branching was quantified as stem biomass invested in side branches relative to biomass of the leader stem: branch mass fraction (BMF) = mass side branches/total mass. Species with a high emphasis on branching have values approaching 1.0, while BMF = 0 for species with no branching.

data considerations

The primary purpose of the current study was to quantify bivariate relationships between species traits (raw data given in Appendix S1 in Supplementary Material) and to compare these relationships among species sampled to span the light and successional gradients. To achieve this we utilized linear scaling (Niklas 1994) relationships between species mean trait values plotted on log scaled axes. Bivariate trait relationships were analysed by fitting standardized major axis (SMA) lines within individual sets, with 95% slope confidence intervals calculated according to Pitman (1939). SMA estimates of the line summarizing the relationship between two variables (i.e. the main axis along which two variables are correlated) are superior to ordinary linear regression estimates for our purposes, because residual variance is minimized in both X and Y dimensions, rather than the Y dimension only (McArdle 1988).

To compare observed bivariate trait relationships among sets, we tested for statistical differences between the slopes and then between the intercepts of set SMA relationships, using (S)MATR 1.0 software (Falster et al. 2003). We tested first for significant heterogeneity among slopes for the two sets, by estimating a common slope (following Warton & Weber 2002) and permuting residuals from the common slope among groups (Manly 1997). Given a common slope (test for heterogeneity not significant), we tested for elevation differences between sets by transforming the data such that the common slope was 0 (Wright et al. 2001) and testing for differences in set means of y′ using one-sample anova (where y′ is y after transformation by an amount corresponding to the common slope β; y′ = y−βy).


potential height

Species ranged in potential height (Hpot) from 4.5 to 42.7 m within the successional set and from 1.04 to 45.2 m within the light gradient set. The asymptotic function (equation 1) provided a significant fit for 39 of 47 species, with r2 > 0.60 for most species and r2 > 0.90 for many species (Appendix S1). The low r2 of some relationships may be attributed to the fact that we did not separate individuals growing in sun vs. shade, or plants with a history of stem damage (Ishii et al. 2000). In the remaining eight species data were insufficient to characterize the asymptotic section of the curve and the predicted Hpot was above that considered reasonable (e.g. > 100 m). For these species Hpot was estimated as the 95th percentile of observed heights.

bivariate trait relationships

There were several significant relationships between Hpot and individual traits within the successional and light-gradient sets. The nature of these relationships (SMA slope and intercept) differed between the two sets. Relationships were weaker or non-significant when all species were pooled (Tables 2 and 3), confirming the importance of separating species between the two sets.

Table 2.  Data for cross-species standardized major axis (SMA) relationships between potential height and other traits fitted within the successional and light gradient sets, corresponding to Fig. 2. Traits were log-transformed prior to analysis. Data given are the range of trait values observed within each set, n, r2 and P-value from test for Pearson correlation between the trait pairs, SMA slope (95% CI) and intercept term. An r2 for a line fitted to all data pooled is also given. Where there was a significant relationship (P < 0.05) within both sets, we tested for significant heterogeneity among set SMA slopes (see text). The test statistic (−2 log Δ) and P-value for this test are given. Where set slopes were not significantly heterogenous (P > 0.05) the estimated common SMA (95% CI) slope is given. Leaf mass area−1 (mg mm−2), LMA; leaf nitrogen mass−1 (mg N mg−1), Nmass; leaf nitrogen area−1 (mg N mm−2), Narea; wood density (mg mm−3), WD; seed mass (mg), SM; branch mass fraction (mg mg−1), BMF. All traits except BMF were log transformed
TraitSuccessional gradient setLight gradient setCommon slope test
Rangenr2PSMAInterceptRangenr2PSMAInterceptr2all−2 log ΔPSMAcom
LMA 0.045–0.149190.57< 0.001 0.60 (0.43,0.83)−1.81 0.055–0.243260.62< 0.001 0.34 (0.26,0.44)−1.330.43 7.15  0.008 
Nmass0.0096–0.035190.12  0.148−0.55 (−0.88,−0.35)−0.99  0.01–0.0237260.00  0.860 0.20 (0.13,0.30)−1.990.00   
Narea 0.001–0.0031190.30  0.016 0.48 (0.32,0.73)−3.360.0008–0.0031260.67< 0.001 0.33 (0.26,0.42)−3.110.53 2.32  0.1300.36 (0.30,0.45)
WD 0.230–0.603190.23  0.039 0.42 (0.27,0.65) −0.94 0.433–0.843260.16  0.040−0.16 (−0.23,−0.11)−0.050.0710.58< 0.001 
SM  0.07–7077170.45  0.003 4.85 (3.26,7.21)−4.97     7–48000120.48  0.013 2.13 (1.31,3.48) 0.230.18 6.57  0.010 
BMF     0–0.68190.01  0.73 0.70 (0.43,1.14)−0.45     0–0.71260.42< 0.001 0.423 (0.31,0.58) 0.00.25   
Table 3.  Data for cross-species standardized major axis (SMA) relationships between potential height and measures of extension cost fitted within the successional and light gradient sets, corresponding to Fig. 3. Traits were log-transformed prior to analysis. Data given are the range of trait values observed, n, r2, P-value from test for Pearson correlation between the trait pairs, SMA slope (95% CI) and intercept term. An r2 for a line fitted to all data pooled is also given. Where there was a significant relationship (P < 0.05) within both sets, we tested for significant heterogeneity among set SMA slopes (see text). The test statistic (−2 log Δ) and P-value for this test are given. Where set slopes were not significantly heterogenous (P > 0.05) the estimated common SMA (95% CI) slope is given. Stem mass length−1 (mg mm−1), SMPL; leaf mass length−1 (mg mm−1), LMPL; leaf mass fraction, LMF (mg mg−1); leaf area ratio (mm2 mg−1), LAR; leaf nitrogen fraction (mg N mg−1), LNF
ScaleTraitSuccessional gradient setLight gradient setCommon slope test
Rangenr2PSMAInterceptRangenr2PSMAInterceptr2all−2 log ΔPSMAcom
Branch tipSMPL  1.28–45.1190.01  0.680 1.60 (0.98,2.61)−1.25   1.5–26.0260.04  0.360 0.67 (0.45,1.00) 0.080.03   
LMPL  1.94–209.9140.03  0.583 1.93 (1.07,3.47)−1.09  3.71–148.9240.01  0.750 0.90 (0.58,1.38) 0.420.05   
LMF  0.57–0.90140.00  0.876 0.22 (0.12,0.40)−0.40  0.53–0.91240.01  0.717 0.14 (0.09,0.22)−0.260.01   
LAR  5.16–16.49140.61  0.001−0.58 (−0.85,−0.39) 1.67  3.56–12.85240.49< 0.001−0.34 (−0.46,−0.25) 1.210.354.690.032 
LNF 0.009–0.029140.14  0.196−0.45 (−0.79,−0.26)−1.20 0.007–0.19240.01  0.702 0.25 (0.16,0.38)−2.150.01   
Terminal meterSMPL  24.8–213.5190.18  0.067 1.04 (0.67,1.63)  0.49  15.8–280.1260.46< 0.001 0.78 (0.58,1.06) 0.980.371.100.2930.86 (0.68,1.09)
LMPL  17.4–271.8190.51< 0.001 1.30 (0.91,1.85)  0.21   7.5–353.5260.74< 0.001 0.98 (0.79,1.21) 0.780.661.890.1741.06 (0.89,1.26)
LMF  0.29–0.74190.47  0.001 0.36 (0.25,0.52)−0.75  0.24–0.66260.55< 0.001 0.22 (0.17,0.29)−0.530.494.420.034 
LAR  4.05–10.77190.19  0.059−0.46 (−0.72,−0.30) 1.35  2.27–8.08260.13  0.072−0.28 (−0.41,−0.19) 0.970.05   
LNF0.0052–0.023190.01  0.688 0.58 (0.36,0.95)−2.740.0038–0.013260.39< 0.001 0.28 (0.20,0.38)−2.370.29   

Leaf traits

There was a strong positive relationship between adult leaf mass per area (LMA) and Hpot in each of the two sets (Fig. 2a, Table 2). Slopes differed significantly between sets (P = 0.008), such that shorter species in the successional set had lower LMA than similar statured species in the light gradient set (Fig. 2a). The relationship across all species pooled was strong (r2 = 0.43), but not as strong as within each set (0.57, 0.62).

Figure 2.

Standardized major axis (SMA) relationships between species’ potential height and (a) leaf mass area−1, (b) leaf nitrogen area−1, (c) wood density, and (d) seed mass. Dashed lines indicate SMA slopes for species in the successional set (open circles), solid lines indicates SMA slopes for species in the light gradient set (closed circles). Slope, r2 and P-values are given in Table 2. In each case we tested for significant differences in slope, then elevation among sets. Set slopes were significantly different for a, c and d, but not for b, although the sets differed significantly in SMA elevation. All axes are log-scaled.

Leaf nitrogen on a mass basis (Nmass) varied from 0.0096 to 0.035 mg mg−1, with the highest values observed among several early successional fast growing species (Appendix S1). Nmass was not correlated with Hpot in either set, or across all species (Table 2). Consequently leaf nitrogen on an area basis (Narea) was correlated with height in a similar manner, as was LMA, in both sets and across all species (Table 2, Fig. 2b). Slopes were not significantly different between sets (P = 0.13). The estimated common slope was 0.36, meaning that across a 10-fold range of height, Narea increased c. 2.3-fold. Individual sets differed significantly in SMA elevation (d.f. = 1, 43, F = 5.32, P = 0.03), with species in the light gradient group having significantly higher leaf nitrogen at a given height.

Wood density

There was a moderate positive relationship between density (calculated using a wood sample taken 1 m back from the tip of a terminal branch) and Hpot in the successional set (r2 = 0.23) and a negative relationship in the light gradient set (r2 = 0.16, Fig. 2c, Table 2, slopes significantly different P < 0.001), but no significant relationship when all species were pooled (r2 = 0.07). The relationship among the successional species was surprisingly loose, with two taller species (Aleurites rockinghamensis, Alstonia scholaris) having low density. These species are known to be fast growing.

Alternative measures of wood density taken from 25 cm back along the terminal branch and at the base of the plant were tightly correlated with the value from 1 m (r2 = 0.82, n = 45, r2 = 0.72, n = 41, respectively). Testing each measure against Hpot yielded similar results for the successional set, but the relationship became non-significant in the light set (r2 = 0.04–0.11, P > 0.05). Consequently, the negative relationship in Fig. 2(c) should be interpreted with caution. This does not alter the overall conclusion that trait relationships were different between the two sets, as slopes of density against Hpot were significantly heterogeneous in each case (P < 0.002, data not shown), with short species in the light gradient having consistently denser wood than short species coming early in succession.

Seed mass

Species mean seed mass varied at least five orders of magnitude within each set (Table 2) and was positively correlated with Hpot across all species (r2 = 0.18, n = 29). However, consistent with our findings for other traits, the nature of the relationship differed between the two sets of species, with tighter correlations observed within individual sets than across the entire species set (Fig. 2d, Table 2). Slopes were significantly different (P = 0.01), with shorter species in the successional set possessing considerably smaller seeds than similar statured species in the light gradient set. Except for a single species (Alstonia scholaris), tall species in each set had similar seed mass values.

Extension costs at tip

Stem mass per unit length (SMPL) was calculated first as the value realized at the tip of the branch. SMPL thereby provides a simple estimate of the biomass cost for a unit of stem extension, excluding any additional thickening further back down the stem. SMPL varied 35-fold among species, but was unrelated to Hpot in both the successional and light gradient sets (Table 3, Fig. 3a). This was contrary to our expectations of a positive relationship in both sets.

Figure 3.

Standardized major axis (SMA) relationships between species’ potential height and (a) stem mass length−1, (b) leaf mass−1 length, (c) leaf mass fraction, (d) leaf area ratio, and (e) leaf nitrogen fraction calculated using samples taken from (left) the branch tip and (right) for a terminal metre of branch (including any additional side branches). Symbols as for Fig. 2. SMA relationships for individual sets are indicated where significant. Slope, r2 and P-values are given in Table 3. In instances where a significant relationship exists within both sets, we tested for significant differences in slope, then elevation among sets. Slopes were significantly heterogenous among species-sets (P < 0.05) for LAR at the tip (d left) and LMF at 1 m (c right), but not for SMPL or LMPL at 1 m (a–b right). All axes are log-scaled.

SMPL at the branch tip is the combination of twig thickness with wood density. Across species, terminal twig cross-sectional area (TwXSA) varied 100-fold while density varied only fourfold. Consequently, density was not a significant predictor of variation in SMPL (Table 4), while TwXSA was tightly correlated with SMPL (Table 4, Fig. 4a). TwXSA was not correlated with Hpot in either set or across all species (data not shown).

Table 4.  Data for cross-species correlations between stem mass length−1, SMPL; leaf mass length−1, LMPL; leaf mass fraction, LMF; leaf area ratio, LAR; and leaf nitrogen fraction, LNF with species traits. Traits were log-transformed prior to analysis. Data given are correlation coefficient (n, P-value from test for Pearson correlation between the trait pairs). Data are not separated with respect to the ecological sets used elsewhere. Here we are interested in how interspecific variation is driven by structural traits independent of patterns related to height
Branch tip
 Leaf mass per area 0.282 (45,0.061) 0.274 (38,0.096) 0.144 (38,0.387)−0.916 (38,0.000)−0.324 (38,0.048)
 Wood density−0.073 (45,0.632)−0.200 (38,0.228)−0.208 (38,0.209)−0.290 (38,0.077)−0.622 (38,0.000)
 Twig cross-section area 0.805 (45,0.000) 0.847 (38,0.000) 0.509 (38,0.001) 0.143 (38,0.392) 0.498 (38,0.001)
 Leaf N(mass)−0.08 (45,0.60) 0.04 (45,0.77) 0.08 (38,0.65) 0.50 (38,0.00) 0.88 (38,0.00)
Terminal metre, including side branches
 Leaf mass per area 0.617 (45,0.000) 0.757 (45,0.000) 0.569 (45,0.000)−0.767 (45,0.000)−0.013 (45,0.935)
 Wood density 0.084 (45,0.582) 0.011 (45,0.945)−0.070 (45,0.649)−0.416 (45,0.004)−0.532 (45,0.000)
 Twig cross-section area 0.318 (45,0.033) 0.262 (45,0.082)−0.007 (45,0.965)−0.032 (45,0.836) 0.284 (45,0.058)
 Leaf Nmass−0.17 (45,0.27)−0.19 (45,0.21)−0.15 (45,0.31) 0.49 (45,0.00) 0.70 (45,0.00)
 Branch mass fraction−0.519 (45,0.000)−0.460 (45,0.001)−0.214 (45,0.157) 0.199 (45,0.189)−0.048 (45,0.755)
 Leaf retention−0.025 (45,0.872) 0.270 (45,0.073) 0.642 (45,0.000) 0.187 (45,0.218) 0.280 (45,0.063)
Figure 4.

Standardized major axis (SMA) relationships between (a) twig cross-sectional area vs. stem mass length−1 (circles, solid line: r2 = 0.65, β = 0.82) and leaf mass length−1 (triangles, dashed line: r2 = 0.72, β = 1.13) at the branch tip, and (b) leaf mass area−1 vs. stem mass length−1 (circles, solid line: r2 = 0.46, β = 2.05) and leaf mass length−1 (triangles, dashed line: r2 = 0.64, β = 2.57) for the terminal metre. Open symbols, species in the successional set; filled symbols, species in the light gradient set. All axes are log-scaled.

Leaf mass per unit length (LMPL) varied across two orders of magnitude (Table 3). Like SMPL, LMPL was tightly correlated with TwXSA (Fig. 4a) and SMPL (r2 = 0.68) but not with Hpot (Table 3, Fig. 3b). Thus species with greater deployment of leaf area incurred a greater cost in stem mass and vice versa. Consequently, there was far less variation in leaf mass fraction (LMF) than was observed for either LMPL or SMPL. LMF at the tip varied twofold across species, compared with 35-fold variation in SMPL. LMF was not correlated with Hpot in either set nor across all species pooled (Table 3, Fig. 3c), but was positively correlated with TwXSA (Table 4). The correlation with TwXSA arose because LMPL increased more rapidly than SMPL with increasing twig cross-sectional area (Fig. 4a).

Partitioning of extension rate into the product of AN, Narea, LMA−1 and LMF (equation 2) showed that variation in stem expenditure had little influence on interspecific differences. Leaf area ratio (LAR = LMF × LMA−1) at the tip varied fivefold across species and was negatively correlated with height in both the light and the successional gradient sets (Table 3, Fig. 3d). LAR was not correlated with TwXSA (Table 4), indicating that the trade-off between LMPL and SMPL leads to little variation in LMF compared with variation in LMA. Slopes of the height-LAR relationships were significantly different, such that short species coming earlier in succession had significantly higher LAR than short late successional species. Interestingly, shorter species in the light gradient set had higher LAR at the shoot tip than tall late successional species (Fig. 3d) despite having higher tissue density. This finding provides strong support for the notion that species maturing in low-light environments maximize LAR through lower LMA.

Combining LAR with Narea gives leaf nitrogen fraction (LNF, mg N mg−1). LNF tested as uncorrelated with Hpot in either set or across all species (Table 3). However, a single short species (Melastoma cyanoides) in the successional set obscured an otherwise tight negative relationship between height and LNF (r2 = 0.40, n = 13, Fig. 3e). Intuitively, a negative relationship fits with our understanding of high extension rates among early successional species and supports our hypothesis (Table 1). Further species would need to be sampled to increase confidence about whether height and LNF were indeed negatively related within the successional set. The lack of a relationship in the light gradient set indicated that high LAR of shorter species was counteracted by the negative relationship between height and Narea (Fig. 2c).

Extension costs for terminal metre

We now apply the same partitioning to data for the terminal metre of branch. Compared with the branch tip, side-branching is expected to play a significant role on allocation profile. Several results were consistent with those observed at the tip: LMPL and SMPL were tightly intercorrelated across species (r2 = 0.79); species with greater LMPL had lower LMF (r2 = 0.41); and overall variation in LMF was small compared with SMPL or LMPL (Table 3).

In contrast to results for the branch tip, LMPL, SMPL and LMF were positively correlated with Hpot within each set and across all species pooled (Table 3, Fig. 3a–c). Except for LMF, slopes with Hpot were not significantly different between sets (Table 3). As LMPL and SMPL covary tightly, we sought to understand the correlation pattern with Hpot by tracing variation in LMPL alone.

Compared with the strength of relationships with LMPL at the tip, TwXSA accounted for little variation in LMPL at 1 m (Table 4). Of the various alternatives, LMA was the trait most closely associated with LMPL for the terminal metre (Table 4, Fig. 4b). High LMA species had a greater total mass of leaf in side branches (r2 = 0.58), but less so along the main stem (r2 = 0.14). Given the tight correlation of LMA with height, LMPL was therefore also correlated with Hpot.

The proportion of biomass in side branches varied from 0 to 71% across species. Compared with LMA, branching index alone accounted for little variation in LMPL (Table 4). We predicted that degree of branching would be positively correlated with height in the successional set (Table 1), but this was not the case (r2 = 0.01, n = 19). In the light gradient set, we hypothesized a negative relationship, but branching index was tightly positively correlated with height (r2 = 0.42, n = 26), such that taller species had a greater proportion of biomass in side branches.

The positive relationship between LMF and Hpot largely cancelled the relationship between Hpot and LMA, such that LAR was only weakly correlated with Hpot in the successional set and uncorrelated in the light set (Fig. 3d). Multiplying by Narea gave a positive relationship between height and leaf nitrogen fraction in the light set, but no relationship in the successional set (Fig. 3e).

Estimated stem extension rates

Equation 2 shows that influences on stem extension rate (SER) within branches can be partitioned into two components, i.e. daily photosynthetic assimilation per mg leaf nitrogen (AN, not measured) and LNF (which gathers together all the properties measured in this study). Thus LNF (Fig. 3e) summarizes the consequences of all the measured architectural and allocation properties.

Based on LNF results at the branch tip, this study indicates a decrease in SER associated with successional status due to combined effects of LMA, wood density and leaf N. However, at the scale of terminal metre, the benefit to short early successional species is lost due to higher LMF of high LMA species, arising from increased leaf retention in side branches. The lack of a significant relationship between LNF and Hpot for the terminal metre was surprising but may be attributed to the unexpected relationship between branching and height. We chose to compare branch segments sampled from mature trees. It is unclear how results for the terminal metre from mature trees compare with patterns observed for saplings, due to potential ontogenetic differences in branching and light environment.


There are two main ways in which cross-species correlations with height might arise (Westoby et al. 2002). First, traits may be correlated due to a physically enforced trade-off, for example between height and whole plant leaf mass fraction. Much progress has been made in recent years outlining the allometric consequences of variation in body size (c. height) for whole plant allocation, resource use and growth (Niklas 1994; Enquist et al. 1999; Enquist 2002). Trait correlations might also arise via strategic association, where combinations of particular trait values are selected for because they give a competitive edge compared with alternative combinations. The many trait combinations observed among species demonstrate that in any given vegetation type, there exist a variety of strategies capable of success in sustaining viable populations.

trade-offs along axes of coexistence

Overall, our results support the idea of two distinct axes of coexistence among short and tall plant species: one operating at a point in time (light gradient), the second arising via temporal, post-disturbance processes (successional gradient). For several stem, leaf and seed traits, the nature of the relationship with height (i.e. slope and elevation of the bivariate plot) differed between the sets of species selected to span these two gradients. These results can be understood in terms of the different strategic trade-offs that allow coexistence along each gradient.

Consider first coexistence along a light gradient at a point in time. There is growing consensus that shorter species succeed via a syndrome of traits promoting slow growth and high survival in low-light environments (King 1996; Thomas & Bazzaz 1999; Sterck et al. 2001; Kohyama et al. 2003; Poorter et al. 2003). Among studies restricted to late successional species, shorter species are characterized by higher wood density (Fig. 2b, King 1991; Thomas 1996; Kohyama et al. 2003), thicker stems, lower LMA (Fig. 2a, Thomas & Bazzaz 1999) and lower photosynthetic efficiency (Thomas & Bazzaz 1999) than taller species. High wood density, in particular, is associated with increased shade tolerance and survival in low light (Augspurger & Kelly 1984; Lawton 1984; Loehle 1988; Osunkoya 1996). Our results also indicate that shorter species achieve higher LAR per unit stem length at the branch tip via low LMA, despite having denser wood. LAR has been shown to be the major determinant of growth rate in low light (Poorter 1999) and is thought to be a key component of a low light growth strategy (Givnish 1988). Recent reviews (Walters & Reich 1999; Reich et al. 2003) suggest the opposite: that LMA is higher among more shade-tolerant species, contributing to a resource-conservation strategy favoured when carbon budgets are marginal (King 1994; Kitajima 1994).

Other work has indicated increased lateral spread of similar sized saplings for shorter species in a light gradient (Sterck et al. 2001; Poorter et al. 2003). Two other studies suggest no relationship (Aiba & Kohyama 1997; Kohyama et al. 2003). Here we report increased branching for taller species, measured for the terminal metre of plant. While we can confidently confirm the absence of branching among several of the shorter shade-tolerant shrubs included in the current study, we are unable to comment on strategies among saplings of taller species as we sampled only mature individuals. The extent to which the current results generalize across different ontogenetic stages and light environments is unknown.

Another way in which shorter species might coexist with taller species is via the timing of recruitment and growth in relation to canopy-opening disturbance. Coexistence along the successional axis is facilitated via trade-offs between height growth rate and stem persistence and shade-tolerance (Huston & Smith 1987; Pacala & Rees 1998). The view that the traits of shorter, early successional species give pre-emptive access to light by allowing early colonization and rapid height growth following disturbance (Coley 1988; Reich et al. 1992) is supported by positive correlations with Hpot for LMA and negative correlations with LAR and LNF at the branch tip and positive correlations with Hpot for wood density and seed mass. A consequence of increased growth potential is reduced survival in low light (Kitajima 1994; Condit et al. 1996; Pacala et al. 1996; Walters & Reich 1999), such that species with the opposite traits are favoured later in succession.

Traits measured in the current study connect directly to processes of dry mass investment within shoots (LMA, tissue density, mass per length) and to recruitment strategy (seed mass). The literature suggests at least two other traits for which contrasting correlations with height should be observed. Several authors have reported a positive correlation between Hpot and stem slenderness (m height/cm d.b.h. measured on equivalent sized saplings) among species sampled along a light gradient (Aiba & Kohyama 1996; King 1996; Sterck et al. 2001; Kohyama et al. 2003; Poorter et al. 2003). Narrow stems facilitate rapid height growth, but are associated with increased mortality due to wind or buckling (Putz et al. 1983). Consequently, early successional species are hypothesized to have more slender stems (Claussen & Maycock 1995; Givnish 1995) leading to a predicted negative correlation with height in the successional set.

The second trait for which contrasting correlations with height may be observed is dry mass gain per unit leaf nitrogen (AN). Thomas & Bazzaz (1999) reported lower AN among shorter species sampled along a light gradient, but hypothesized the opposite or no relationship for species in the successional set, reflecting genetic adaptations to different adult light environments. Such a pattern would amplify results of the current study by increasing the separation among sets observed for LNF (Fig. 3e), as LNF and AN multiply to give estimated stem extension rate.

other sources of variation

The findings of the current study give some promise for furthering our understanding of trait correlations with height. Several relationships, however, were weaker than expected (low r2), while others were not significant. One reason is that strategic trade-offs between individual traits might inherently be loose (e.g. Thomas & Bazzaz 1999; Kohyama et al. 2003; Poorter et al. 2003), at least compared with physically enforced relationships (Westoby et al. 2002). A second possibility is that height-related trade-offs only partly account for variation in other traits and that additional factors such as hydrology or soil nutrients must also be accounted for. A third reason may be that the true relationships are partially obscured by biases or limitations in our methods.

Species in the present study occurred in different light, wind and humidity environments. Traits such as LMA, Narea, lateral spread and wood density are known to vary with light level up to two- or threefold within species, in a manner consistent with the relationships reported here (Ellsworth & Reich 1993; Poorter et al. 1995; Sterck & Bongers 2001; Osada et al. 2002). Branching appears particularly sensitive to light environment and ontogeny, which might explain the surprising results obtained for allocation profile in the terminal metre. On the other hand, LMA, Narea and wood density are known to vary by more than twofold across species (e.g. Ter Steege & Hammond 2001; Wright et al. 2004) with four- to sixfold observed in the current study. We believe differences between species would have been apparent even if it had been possible to measure mature individuals in a common light environment. In any event, the differences measured in this study were those that actually occur in the field, including phenotypic effects and genotypic differences and the effects of genotypes in selecting environments.

Variation in other traits may simply be unrelated to patterns of height-growth. For example, the basic cost of stem extension at the branch tip (ranging 30-fold across species) was uncorrelated with potential height in both sets and across all species pooled, despite intuitive arguments for some association. Fast growing early successional species had been expected to economise on stem expenditure per unit length and species high in the light profile to require additional reinforcement to withstand increased exposure to high wind (Givnish 1995). Twig cross-sectional area (and its correlate leaf size) was responsible for most differences in extension cost at the tip and was not associated with height.

potential height as an indicator of ecological strategy

Potential height has been proposed as one of several traits capable of summarizing substantial variation among species’ ecologies (Westoby 1998; Weiher et al. 1999; Westoby et al. 2002). Our results, however, confirm the view that potential height alone cannot fully express a species’ strategy for light capture, because it does not distinguish between the contrasting strategies in early vs. late successional situations (see also Westoby 1998; Thomas & Bazzaz 1999; Sterck et al. 2001; Poorter et al. 2003). This distinction might be made by considering combinations of potential height with LMA, wood density or seed mass (Fig. 2). LMA seems less useful in this respect, as the separation between sets was too small for species to be partitioned without a priori knowledge of the species’ ecology. A benefit of seed mass is that values are already known for large numbers of species around the globe (Moles et al. 2004). Variation in wood density on the other hand connects more directly to processes of growth (Niklas 1994; Castro-Diez et al. 1998; Enquist et al. 1999), mortality (Augspurger & Kelly 1984) and hydraulic conductance (Hacke et al. 2000; Hacke et al. 2001).

How else might we define height strategy? Existing data indicate an increase in shade tolerance and decrease in growth rate with time since disturbance (Pacala et al. 1996), with a similar trend observed down the light gradient (Thomas & Bazzaz 1999; Kohyama et al. 2003). Consequently, the two axes of potential height (Fig. 1) may correspond to a single axis defined by successional status, with short early successional species at one end of the spectrum, tall late successional trees in the middle and short late successional shrubs at the highly shade-tolerant end of the spectrum. A comprehensive test of this proposition would be to attempt to demonstrate a trade-off between height growth rate and survival in low light, independent of differences in potential height. Our data suggest WD and Nmass may be useful indicators of a species position along the axis, as both were correlated with leaf nitrogen fraction, our indicator of stem extension rate (Table 4). Adopting either trait in strategy schemes as an indicator of light capture strategy would correspond to a shift in emphasis from potential height to height-growth rate as the ecological outcome giving pre-emptive access to light.


Warm thanks to W. Edwards, A. Graham, D. Hilbert, P. Juniper & J. Wells for advice and assistance. D. Hilbert from CSIRO TFRC in Atherton generously supplied height-diameter for many species. S. Burchill, D. Falster, R. Jensen, R. Khoury, B. Krug, J. McDowell, A. Moles and B. Rice helped in the field and laboratory. J. Bragg, J. Wells, D. Sheil, D. Burslem and three anonymous reviewers provided valuable comments on earlier drafts. This work was supported by Australian Research Council funding to MW and a Macquarie University PGRF to DF.

Supplementary material

The following material is available from

Appendix S1 Mean trait values for 45 woody shrub/tree species from Queensland's wet tropics.