•Leaf toughness is thought to enhance physical defense and leaf lifespan. Here, we evaluated the relative importance of tissue-level leaf traits vs lamina thickness, as well as their ontogenetic changes, for structure-level leaf toughness and regeneration ecology of 19 tropical tree species.
•We measured the fracture toughness of the laminas and veins of sapling leaves with shearing tests, and used principal component analysis and structural equation modeling to evaluate the multivariate relationships among traits that contribute to leaf toughness and their links to ecological performance traits.
•Tissue traits (density and fracture toughness of lamina and vein) were correlated positively with each other, but independent of lamina thickness. The tissue traits and lamina thickness contributed additively to the structure-level toughness (leaf mass per area and work-to-shear). Species with dense and tough leaves as saplings also had dense and tough leaves as seedlings and adults. The patterns of ontogenetic change in trait values differed between the seedling-to-sapling and sapling-to-adult transitions.
•The fracture toughness and tissue density of laminas and veins, but not the lamina thickness, were correlated positively with leaf lifespan and sapling survival, and negatively with herbivory rate and sapling regeneration light requirements, indicating the importance of tissue-level leaf traits.
Leaf toughness has been defined and measured differently by different researchers using shearing, punching and tearing tests (Read & Sanson, 2003). Fortunately, results are comparable when measurements are normalized properly (Aranwela et al., 1999; Onoda et al., 2008), either per unit fracture area (tissue-level measurement) or per unit fracture length (structure-level measurement that takes the size of the structure into account). For example, fracture toughness (also known as specific work-to-shear, J m−2) is measured as the work per unit cross-sectional area when a lamina or vein is sheared with a blade (Lucas et al., 2000; Read & Sanson, 2003). By contrast, lamina work-to-shear (Ws, J m−1) represents the fracture resistance normalized per unit fracture length, and relates to lamina fracture toughness (LamFT, J m−2) and lamina thickness (LamT, m) as follows:
Likewise, leaf mass per area (LMA, g m−2) is a product of the lamina density (LamD, g m−3) and LamT (m) as:
Many ecological studies have compared leaf mechanical strength at the structure-level only (e.g. Coley, 1983; Poorter, 2009). Decomposing structure-level traits, Ws and LMA, to tissue-level properties and LamT elucidates the nature of the physical defenses of leaves. For example, the high LMA of inherently slow-growing species from light- or nutrient-limited habitats is mainly attributable to high tissue density rather than the thickness of leaves (Shipley et al., 2006; Poorter et al., 2009). Among eight Neotropical tree species, seedling survival has been found to be correlated positively with high fracture toughness and tissue density, but not with LamT (Alvarez-Clare & Kitajima, 2007). This study also found that tough and dense lamina tissues were associated with tough and dense veins and stems, indicating overall concordance of multiple mechanical strength traits at the leaf and whole-plant levels. Decomposing structure-level leaf toughness to tissue-level toughness and LamT is important because a leaf may be thick solely as a result of thick mesophyll tissues (Onoda et al., 2008), and such leaves may still be heavily attacked by herbivores and pathogens.
Our conceptual framework is summarized graphically in Fig. 1(a), which shows how among-species’ variation in matter distribution at the structural level (LMA) can be partitioned to variations in LamD (tissue trait) and LamT (size trait), and how species’ differences in Ws (structure-level leaf toughness) can be attributable to the variation in LamFT (tissue trait) and LamT (size trait). We hypothesize that tissue density, which largely reflects cell wall dry mass per unit volume space (Poorter et al., 2009), should be correlated strongly with fracture toughness, based on the findings by Choong et al. (1992) and Alvarez-Clare & Kitajima (2007). Most importantly, we predict that these leaf toughness-related traits influence ecological performance traits (Fig. 1a). If so, they contribute to the functional basis for growth–survival trade-offs found among coexisting tree species (Kitajima, 1994; Kobe, 1999; Walters & Reich, 1999; Wright et al., 2003; Gilbert et al., 2006; Poorter & Bongers, 2006).
In this study we attempt to understand how these components of leaf mechanical strength vary among species in relation to ecological performance traits, including herbivore resistance, leaf lifespan, survival and growth rates of individuals, and regeneration light requirements. We quantified the morphological and biomechanical traits of sapling leaves for 19 species in a tropical moist forest in Bolivia to ask the following questions.
•What is the relative importance of tissue traits (density and fracture toughness) vs size traits (LamT and lamina area) in explaining among-species’ variation in leaf mechanical traits at the structure-level (LMA and work-to-shear)?
•How do the morphological and biomechanical traits of juvenile leaves correlate with the ecological performance of leaves (palatability to a generalist herbivore, leaf lifespan) and whole plants (survival and growth rates), as well as the regeneration light requirements of the species?
•How are leaf biomechanical and morphological traits of saplings correlated with those of earlier and later ontogenetic stages?
We predicted that the interspecific variation in sapling leaf toughness at the structural level should reflect species’ differences in tissue-level properties as well as LamT (Fig. 1a). We also predicted that shade-tolerant species with higher leaf and plant survival rates should exhibit greater leaf toughness at both the tissue and structure-levels. Following the findings of Alvarez-Clare & Kitajima (2007) for seedlings, fracture toughness and tissue density, rather than LamT, were expected to show positive correlations with leaf and sapling survival. We evaluated these multivariate causal hypotheses by structural equation modeling, which offers a powerful approach for the integration of functional traits and their ecological significance (Shipley, 2000). Leaf size is known to vary dramatically among ontogenetic stages, as well as among species (Ackerly & Donoghue, 1998; Westoby & Wright, 2006; Poorter & Rozendaal, 2008). Hence, our a priori prediction was that large leaves would require greater support at the structure-level by increasing LamT and/or increasing tissue-level toughness. To our knowledge, this is the first study to examine species’ differences in mechanical and morphological leaf traits at multiple ontogenetic stages of tree species, linking them directly to their ecological performance.
Materials and Methods
Research site and species’ selection
Field work was carried out in the lowland tropical moist semi-evergreen forest of La Chonta (15°47′S, 62°55′W), Bolivia, a long-term research site of the Instituto Boliviano de Investigación Forestal, at an altitude between 300 and 500 m. Annual precipitation in La Chonta is 1580 mm, with a dry season (< 100 mm per month) from April until October. The forest has an average canopy height of 27 m, stem density of 367 ha−1, basal area of 19.3 m2 ha−1 and species’ richness of 59 species ha−1 (all data for trees > 10 cm in diameter at breast height (DBH); Pena-Claros et al., 2008).
We selected 19 common tree species (Table 1), 16 of which reach DBH > 10 cm, comprising 45% of the tree stems at the site. The remaining three species were small subcanopy trees. These species differed in regeneration guilds and other ecological characteristics (Table 1), including regeneration light requirements, leaf lifespan, sapling growth and survival (Poorter & Bongers, 2006), wood density (van Gelder et al., 2006) and maximum adult height (Poorter et al., 2006). We also used an independent, objective and continuous index of the regeneration light requirements of the species (i.e. the inverse of shade tolerance) by analyzing the crown exposure (CE) in relation to the height of individual trees for each species (Poorter & Kitajima, 2007). To this end, on average, 523 individuals per species (median, 263; range, 41–9319) were sampled over their whole size range for the measurement of height and CE (Dawkins & Field, 1978). CE is scored as follows: 1, if the tree does not receive any direct light; 2, if the tree receives lateral light; 3, if the tree receives overhead light on part of the crown (10–90% of the vertical projection of the crown exposed to full overhead light); 4, if the tree receives full overhead light (> 90% of the vertical projection of the crown exposed to vertical light); and 5, if the tree has an emergent crown that has no obstructing foliage over an upward facing cone encompassing the crown, with angles of 45° from the horizontal). For each species, CE was regressed against tree height using a multinomial regression analysis (cf. Poorter et al., 2005; Sheil et al., 2006), and the average population-level CE at a standardized height of 2 m (juvenile CE, CEjuv) was calculated. Although similar-sized individuals of the same species can be found under a wide range of CEs, we were interested in estimating a population average CEjuv as an index of the regeneration light requirement of the species (Poorter et al., 2005).
Table 1. Study species, their ecological characteristics (means for: Hmax, maximum adult height; Guild, regeneration guild; CEjuv, average crown exposure of 2 m tall saplings) and key leaf traits of saplings quantified in this study (LamFT, lamina fracture toughness; LamT, lamina thickness; LMA, leaf mass per area measured with 12.9 mm diameter disks; Ws, work-to-shear of lamina)
Leaf lifespan (months)
Lamina area (cm2)
LMA (g m−2)
LamFT (J m−2)
Ws (mJ m−1)
The species are listed in decreasing order of leaf lifespan. Asterisks indicate the 11 species for which leaves of greenhouse-grown seedlings were also sampled.
For the determination of sapling leaf traits, we sampled typical mature leaves from saplings of each species in May 2005, at the beginning of the dry season when the humidity and soil water availability were still high. At least four individuals per species were sampled for 17 species, whereas the sample size was three for two species, which is considered to be sufficient for leaf physical strength measurements (Cornelissen et al., 2003). For each leaf, we measured the fresh mass and traced the leaf shape on a sheet of paper immediately following collection and removal of the petiole. The leaves were kept in a plastic bag with a small piece of moist filter paper until shearing tests. The shearing tests (see Biomechanical measurements section) were conducted within 4 h of collection. Adult leaves were sampled from trees with CE ≥ 3 and DBH ≥ 10 cm, choosing leaves from the outer leaf layer midway between the top and bottom of the crown (Rozendaal et al., 2006). To determine lamina-specific LMA and LamD, two circular disks of 12.9 mm in diameter were collected with a cork borer from each leaf, excluding primary and secondary veins. LamT was measured with a micrometer (SM112; Teclock, Okaya, Japan), avoiding any visible veins. After measurements with fresh leaves, leaves and leaf disks were oven dried at 65°C for a minimum of 3 d to determine the dry mass. LamD was calculated by dividing LMA by LamT. LMA, LamD and the leaf dry matter content (LDMC = ratio of dry mass to fresh mass) were highly correlated between the whole blade and circular disk determinations (Pearson r =0.99, P <0.001 for all three variables; see Table 1 and Supporting Information Table S3 for species’ means); thus, we report only the results using values calculated from the circular disks.
Leaves from a minimum of four seedlings were sampled for 11 species (Table 1). The seedlings were grown from seeds in a greenhouse in Santa Cruz, Bolivia (16°30′S, 68°10′W), in 200 ml plastic tubes filled with a 50 : 50 mix of sand and forest soil under 10% of daylight (see Poorter & Markesteijn (2008) for further details). Seedlings were c. 2 yr old at the time of this study. As LMA values of greenhouse-grown seedlings are a good approximation of the LMA values of field-sampled seedlings (Bloor, 2003), leaf traits of greenhouse seedlings were likely to approximate seedling leaf traits at the same ontogenetic stage under a similar light environment in the field.
We measured the work-to-shear of leaves of seedlings and saplings using a portable universal tester to which a pair of scissors was attached (described in Darvell et al. (1996)), following the method described by Lucas et al. (2001). A small rectangular section, c. 8–9 mm wide and 15–18 mm long, was cut from the central portion of the leaf blade, so that the central vein bordered a long edge of the rectangle. The rectangular section was sheared with the scissors in a single path perpendicular to the central vein, avoiding secondary and tertiary veins in the lamina portion. The force–displacement curve (i.e. shearing force recorded with a load cell plotted against displacement distance) was distinctive for shearing the vein and lamina tissues, from which the total of shearing each tissue type was determined. The total of shearing the lamina was divided by the rectangular cross-sectional area (=fracture length × LamT) to calculate LamFT. The total work to shear of the vein was divided by the vein cross-sectional area (=πr2) to calculate the vein fracture toughness. The lamina work-to-shear (Ws) was calculated by multiplying LamFT by LamT. When the central vein in the leaf center was wider than 2.5 mm, a rectangular area further towards the tip (where the diameter was between 1.8 and 2.5 mm) was selected for the test. Because the vein was not cut at the same position for all species, we did not analyze the structure-level toughness of the veins (i.e. vein work-to-shear, which is dependent on the diameter). The vein used for the cutting test was separated from the lamina with a razor blade to determine the vein density (VeinD) after drying at 65°C.
For adult leaves of 18 species and sapling leaves of 13 species, we measured the force-to-punch (Fp) using a penetrometer constructed with a punch rod with a diameter of 3 mm, mounted on a stationary base with a hole (Rozendaal et al., 2006; Poorter, 2009; L Poorter, unpublished). The leaf was positioned to avoid primary and secondary veins, and the maximum force registered was divided by the rod circumference to normalize per fracture length and to determine Fp. Fp was subsequently divided by LamT to calculate the specific force-to-punch (Fps). Across species, the results of the shear test and the punch test were correlated linearly with each other (Supporting Information Fig. S1); Fps was correlated with LamFT (tissue-level correlation, r =0.65, P =0.013, n =18), whereas Fp was correlated with Ws (structure-level correlation, r =0.86, P <0.001, n =18).
For 18 species, palatability to a generalist herbivore was measured as part of a larger experiment with 40 species. Fresh leaves were collected to test the palatability to a snail species in a cafeteria experiment (cf. Cornelissen et al., 1999). A specialist at the Universidad Nacional de Tucumán, Argentina, was consulted for the taxonomic identification of this locally common snail species. Four square pieces of leaf (1 × 1 cm2) per species were placed at random positions in a plastic tray (42 × 42 cm2) to offer feeding choices to 15 snails per tray. The experiment was replicated three times. The percentage of lamina area remaining was quantified six times during a 5-d period, from which the loss rate per hour was calculated as the absolute value of the slope of log10(% area remaining + 1) plotted against time (0–96 h). To achieve normality of the distribution, the loss rate was log10 transformed after adding a small number (0.000001, which is 1/100 of the lowest nonzero loss rate), and this value was used as the ‘palatability index’, which ranged from −7 for species that were not consumed at all (slope = 0) to −0.83 for Urera caracasana which was completely consumed within 72 h. The palatability index was correlated positively with the observed field herbivory rates (% lamina area loss per month) for 12 species whose sapling leaves were marked and monitored in the forest as part of a study with 30 species (L Poorter, unpublished; r =0.58, P <0.05, Table S3). The two measures of herbivory exhibited similar trends in relation to leaf traits measured in the study. Thus, we report only the analyses using the palatability index determined for the larger number of species.
All measured traits varied significantly among species (one-way ANOVA, results not shown). Tests with two-way ANOVA indicated that both species and ontogeny affected significantly all traits measured for multiple ontogenetic stages (results not shown). Hence, in all subsequent analyses, species’ mean values were used at each ontogenetic stage, after log10 transformations for variables that did not meet the assumption of normality. The survival rate was the only variable that remained non-normal with log or arcsine-square root transformations, and thus correlation analysis involving survival rate was examined using Spearman rank correlation. All other associations between pairs of trait variables were examined using Pearson correlation and standard major axis (SMA) regression with SMATR software (Warton et al., 2006). SMA regression slopes were compared to test whether the relationship changed significantly between ontogenetic stages in terms of regression slope, elevation and shifts along the regression line (Warton et al., 2006). Multivariate associations among lamina and vein traits relevant for mechanical strength and size of sapling leaves were summarized using principal component analysis (PCA). Then, we examined the correlations of the first three PCA axes with sapling traits directly relevant for the regeneration strategies of the species, including palatability index, leaf lifespan, growth rates, survival rates, CEjuv and wood density. The hypothetical causal structure (Fig. 1a) was evaluated using structural equation modeling (Shipley, 2000; Grace, 2006) with AMOS software (AMOS Development Corporation, Spring House, PA, USA). We compared alternative path models by evaluating the statistical significance of individual paths and the overall model fitting using maximum likelihood chi-squared and the Akaike Information Criterion (AIC). We also used the orthogonal trait axes for leaf traits extracted by PCA as exogenous variables that might influence leaf lifespan, survival, growth rates and light requirements of saplings. Phylogenetic independent contrasts and their correlations were calculated using Phylocom software, following Webb et al. (2008), as described in Supporting Information Table S1. All statistical analyses were conducted using JMP software (SAS Institute, Cary, NC, USA) unless otherwise stated.
Mechanical strength of sapling leaves and its determinants
The mechanical strength of sapling leaf lamina varied widely among the 19 tree species (Table 1); LamFT (tissue-level toughness) differed ninefold among species (53–476 J m−2), whereas Ws (structure-level toughness) varied 12-fold (10–123 mJ m−1) and LamT (size trait) varied less than threefold among species (0.14–0.34 mm). The mechanical strength at the tissue and structure-levels reflected matter distribution within the same level. At the tissue level, LamFT was correlated positively with measures of dry mass density, LamD (Fig. 2a) and LDMC (Table S1). At the structural level, Ws was correlated positively with LMA (Fig. 2b). The proposed path model among components of leaf toughness (Fig. 1a) was well supported (Fig. 1b), indicating that the variation in LamFT (β = 0.85, P <0.001) contributed much more strongly than that of LamT (β = 0.49, P <0.05, Fig. 1b) to the variation in Ws. Lamina and vein fracture toughness values were correlated strongly with each other across species, even though veins were c. 10 times tougher than laminas in all species (Fig. 3a). Unexpectedly, the fracture toughness values of laminas and veins were correlated negatively with the lamina area, indicating that species with large leaves had weak lamina tissues (Fig. 3b). These pair-wise correlations between leaf traits were similar when phylogeny was taken into account (Supporting Information Table S2).
A PCA (Table 2, Supporting Information Fig. S2) revealed that tissue-level leaf traits varied independently of LamT. The fracture toughness of laminas and veins, as well as their dry mass density, loaded positively on PCA1, which explained 65% of the variation, but showed weak or negative correlation with PCA2, which explained an additional 21% of the variation (Table 2). By contrast, LamT contributed little to PCA1, but was the strongest contributor to PCA2. The two structure-level traits that should reflect both tissue properties and LamT (i.e. LMA and Ws) showed similar degrees of loading on PCA1 and PCA2 (Table 2). Lamina area was the single strongest contributor to PCA3, which explained 6% of the variance, in addition to the 86% explained by the first two axes. In summary, PCA1 captures the variation in dry mass density and tissue-level mechanical strength, whereas PCA2 and PCA3 summarize two aspects of leaf size, i.e. LamT and lamina area, respectively. The results were identical when all variables were log transformed first (i.e. the same variables contributed to PCA1–3, and together explained 93% of the variance).
Table 2. Principal component analysis (PCA) for lamina and vein traits relevant for leaf toughness, showing the eigenvector values on each of the first three principal component axes
The traits are listed in the order of their contribution (eigenvector value) to the first principal component axis (PCA1).
LDMC, leaf dry mass/fresh mass; LMA, leaf mass per area.
Vein fracture toughness
Lamina fracture toughness
log10 (lamina area)
Relationships of leaf mechanical traits to ecological performance traits
The ecological performance traits of the 19 Bolivian tree species exhibited strong correlations with both tissue- and structure-level toughness in the predicted directions, but exhibited no correlations with lamina area and thickness. Species with tougher and denser leaves exhibited lower palatability, longer leaf lifespan, higher sapling survival, slower height growth rates and lower CEjuv (Fig. 4, Table 3). LamFT and PCA1, which summarize tissue toughness and density, showed identical correlations with these ecological performance traits (Fig. 4). The same results were found when phylogeny was taken into account (Table S2).
Table 3. Pair-wise correlations of lamina and vein traits relevant for leaf biomechanical strength with the ecological performance of saplings, regeneration light requirement (CEjuv) and wood density of saplings, a stem functional trait
Sapling survival rate
Pearson correlation coefficients (bold if P <0.05) are shown for all variables, except that Spearman rank correlation coefficients are shown for sapling survival rate which could not be corrected for normality even with transformation (P values in parentheses). See Table 2 for the relative contribution of variables to PCA1.
LDMC, leaf dry mass/fresh mass; LMA, leaf mass per area.
Lamina fracture toughness
0.88 (< 0.0001)
0.78 (< 0.0001)
Vein fracture toughness
0.82 (< 0.0001)
0.87 (< 0.0001)
0.81 (< 0.0001)
0.79 (< 0.0001)
0.87 (< 0.0001)
0.90 (< 0.0001)
As expected, mechanically stronger leaves were less likely to be eaten, and more likely to live longer. Palatability by snails was correlated negatively with all leaf traits that contributed to leaf mechanical strength (Table 3, Fig. 4a,b), with fracture toughness and Ws showing stronger correlation than tissue density and LMA. Likewise, sapling leaf lifespan showed strong positive correlations with LamFT and other traits associated with it (Fig. 4b,c, Table 3). By contrast, LamT contributed little to species’ variation in palatability and leaf lifespan (Table 3).
Leaf mechanical properties and matter distribution were also associated strongly with the survival and growth rates of saplings, as well as species’ light requirements. Species with higher LamFT survived better (Fig. 4e) even though they tended to grow more slowly (Table 3). The negative correlation of CEjuv with tissue density and toughness of laminas and veins (Table 3) indicates that shade-tolerant species with low CEjuv have dense and tough leaf tissues. The causal hypothesis depicted as the path model (Fig. 1a) was supported by the data (Fig. 1b), which showed that LamFT contributed much more strongly than Ws to CEjuv in terms of both direct and total effects (see Supporting Information Fig. S3 for details including the evaluation of alternative path models). The PCA1 score, which summarizes tissue-level toughness and matter distribution, exhibited positive and negative relationships with sapling survival and growth (Fig. 4f,h) in similar manners as LamFT did (Fig. 4e,g, Table 3). By contrast, PCA2 and PCA3 scores, which largely reflect leaf thickness and lamina area, exhibited no significant pair-wise correlations with any ecological performance traits (P >0.2).
Ontogenetic correlations and shifts of leaf biomechanical traits
Species’ means of many leaf traits changed during seedling-to-sapling and sapling-to-adult transitions, but species’ ranks in these traits were concordant across ontogenetic stages. Species with tough and dense leaves as seedlings had tough and dense leaves as saplings (n =11 species, r =0.59–0.91, Table 4), and those with tough and dense leaves as saplings had tough and dense leaves as adults (n =18 species, r =0.72–0.89, Table 4) (see Supporting Information Table S4 for species’ means for seedlings and adults). Tissue density, LMA and LamT were also correlated positively between ontogenetic stages. Species’ ranks in leaf traits were therefore concordant across ontogenetic stages (Table 4). Lamina area was the only trait that did not exhibit significant ontogenetic concordance (r =0.01 for the seedling-to-sapling transition and r =0.45 for the sapling-to-adult transition, Table 4).
Table 4. Ontogenetic shifts of species’ mean values for lamina and vein traits relevant for leaf toughness from seedlings to saplings and from saplings to adults
n, number of species. A positive paired t-value means that trait values increase with ontogenetic stages at the indicated P level with a two-tailed test. Also shown are Pearson correlation coefficients (r) for ontogenetic concordance between the two stages. ns, P ≫ 0.05; na, data not available. LMA, leaf mass per area.
aTissue-level toughness was quantified as the fracture toughness determined by the shearing test for seedlings and saplings, and as specific force-to-punch determined with a penetrometer for saplings and adults. bStructure-level toughness was quantified as work-to-shear for seedlings and saplings, and as force-to-punch for saplings and adults.
Lamina tissue density
Lamina tissue-level toughnessa
Vein tissue-level toughnessa
Lamina structure-level toughnessb
Many species showed significantly increased LMA and structure-level toughness (Ws and Fp) with ontogenetic development (P <0.05 according to paired t-test, Table 4), but the mechanistic reasons underlining these structure-level increases differed between the two transitions. In the seedling-to-sapling transition, there was no general pattern for changes in tissue density and toughness across species (P >0.05, paired t-test, Table 4); when sapling LamFT was plotted against seedling LamFT (Fig. 5a), most points fell near the 1 : 1 line. By contrast, LamT increased significantly from seedlings to saplings (with all points above the 1 : 1 line in Fig. 5c; paired t-test P <0.001, Table 4). Thus, the increase in Ws from seedlings to saplings (Table 4) was solely a result of an increase in LamT (Fig. 5c). By contrast, during the sapling-to-adult transition, Fps (Fig. 5b) and LamD increased significantly (paired t-tests, Table 4), whereas LamT actually decreased in many species (Fig. 5d). Thus, significant increases in structure-level traits, LMA and Fp, from saplings to adults were solely a result of the increases in LamD and tissue toughness.
The lamina area increased significantly from seedlings to saplings (paired t-test, P <0.001), but the degrees and directions of the changes differed so much among species that there was no interspecific correlation in lamina area between seedling and sapling stages (r =0.01, P =0.97, Table 4). During the sapling-to-adult transition, there was neither a general pattern of lamina area change across species (paired t = −0.6, P =0.55) nor significant ontogenetic concordance (r =0.45, P =0.059, Table 4). In summary, ontogenetic shifts of individual leaf traits may switch directions between the seedling-to-sapling transition and the sapling-to-adult transition (Fig. 6), even though overall species’ ranks were maintained across ontogenetic stages (with the exception of the lamina area).
Effects of ontogeny on trait-to-trait and trait-to-performance relationships
The relationship between fracture toughness and tissue density was identical for seedlings and saplings (Fig. 2a). By contrast, the relationship between two structure-level traits, Ws and LMA, differed slightly between seedling and sapling stages, showing that seedlings tend to have higher Ws at a given LMA (Fig. 2b). LamT was independent of tissue-level traits at all ontogenetic stages (P >0.05), and ontogenetic changes in lamina area were not necessarily accompanied by changes in LamT.
The relationships of lamina toughness with performance traits were maintained across ontogenetic stages, indicating that leaf toughness values of seedlings and adults can predict the ecological performance of saplings. Seedling LamFT was correlated positively with sapling leaf lifespan (n =11 species, r =0.80, P =0.003) and survival (n =11 species, rs= 0.63, P =0.038). Adult lamina Fp was correlated with the palatability index (n = 16 species, r = −0.68, P = 0.002), leaf lifespan (n =18 species, r =0.74, P < 0.001), survival (n =18, rs= 0.50, P =0.03) and height growth rate (n =18 species, r = −0.57, P =0.013) of saplings.
The results of this study strongly suggest that tissue-level leaf toughness is the essence of the link between structure-level leaf toughness and the ecological performance of juvenile trees. The same conclusion was reached by Grubb et al. (2008) in their reanalysis of the 46-species dataset of Coley (1983), as well as in a study of young seedlings of eight Panamanian tree species by Alvarez-Clare & Kitajima (2007). Our study supports this conclusion more convincingly by demonstrating the link between leaf mechanical strength and multiple measures of ecological performance and the regeneration light requirements of tree species.
Tissue density and toughness as the functional basis of tough leaves
LMA is widely recognized as a trait that correlates positively with structure-level leaf toughness, such as Ws (Wright & Westoby, 2002; Read & Sanson, 2003; Hanley et al., 2007; Mediavilla et al., 2008), and negatively with photosynthetic rates and nitrogen contents per unit mass (Wright et al., 2004). These correlations exist because the contents of metabolically active mesophyll cells, rather than cell walls and cuticles, occupy large proportions of leaf volume space in species with low LMA (Niinemets, 1999; Shipley et al., 2006). We found that LMA predicted Ws (r =0.87), although LamFT alone predicted Ws equally well (r =0.87, Table S1). The evaluation of the proposed path model (Fig. 1a) demonstrated that tissue density and toughness played a greater role than LamT in linking LMA to Ws (Figs 1b, S3).
Surprisingly, fracture toughness values of both laminas and veins were correlated negatively with lamina area among the study species (Fig. 3b). This was the opposite of our a priori expectation that larger leaves should require greater mechanical support to maintain optimal leaf display. LamT did not compensate for weak laminas of large leaves, because LamT and lamina area were not correlated with each other (r =0.08, Table S1). However, the central vein diameter was correlated positively with leaf size (r =0.80, P <0.01; log10(vein diameter) = 0.35 log10(lamina area)−0.67). The four species with the largest leaves were light-demanding pioneers of Urera, Heliocarpus, Cecropia and Jacaratia. Visual observations suggest that these large-leaved species employ ‘kite’-type leaves common among pioneers (Grubb, 1986), in which thin and soft laminas are supported by large-diameter veins. These mesophyll-rich large leaves may also depend on turgor for self-support, as suggested by frequent wilting later in the day.
Importance of tissue density and fracture toughness for ecological strategies of trees
The results of this study strongly suggest that the tissue density and toughness of laminas and veins are important as part of the shade tolerance strategy characterized by a long leaf lifespan, high survival rates and slow growth rates during the regeneration phase (Kitajima & Poorter, 2008). The optimal leaf lifespan should be longer in shady and other resource-limited environments, where the carbon cost of leaf construction is high relative to the potential photosynthetic income (Coley et al., 1985; Williams et al., 1989; Kikuzawa, 1991). In general, shade-tolerant evergreen species tend to have long-lived leaves with high LMA, low nitrogen contents and low photosynthetic capacity per unit mass (Walters & Reich, 2000; Niinemets, 2006; Poorter & Bongers, 2006; Lusk et al., 2008). Compiling data for 597 woody species, Niinemets (1999) showed that leaf tissue density, rather than LamT, underpins this trade-off between photosynthetic capacity per unit mass and LMA. Parallel to this, the results reported here demonstrate that tissue density, rather than LamT, is the aspect of LMA that enables leaves to achieve a long lifespan against biotic hazards. The central role of tissue density and toughness for physical defense is evident in the slower rates of consumption of tougher leaves by a generalist snail (Fig. 4) and the community of herbivores in the field (n =14 species, r = −0.74, P <0.003, L Poorter, unpublished).
The negative correlation between components of tissue toughness and CEjuv (Table 3, Fig. 4g,h) supports directly the predicted link of physical defense with regeneration light requirements. By contrast, LamT and leaf size showed no correlation with CEjuv. A previous study also found that CEjuv was correlated with punch strength (r = −0.46, P =0.004) and tissue density (r = −0.53, P =0.001) of lamina, but not with LamT (r = −0.05, P =0.75), for adult leaves of 39 tree species at the same site (sampled from 10–20 cm DBH individuals; Poorter, 2009). The associations of leaf toughness-related traits with ecological performance are evolutionarily convergent, as phylogenetic independent contrasts show trait correlations that are of a similar strength to the nonphylogenetic cross-species’ correlations (Table S2).
Functional integration of traits with structural equation modeling
The predicted causal links among traits that contribute to leaf mechanical strength, ecological performance and regeneration strategy are well supported by structural equation modeling using the results of this study. The best-supported model among the alternatives (Figs 7, S3) showed that a higher score of PCA1 (an integrative trait index representing high tissue density and toughness of laminas and veins, and smaller lamina area, Table 2) indicated a longer leaf lifespan and better sapling survival (via leaf lifespan), but slower sapling growth rates (Fig. 7). By contrast, PCA2 (reflecting LamT and Ws) and PCA3 (reflecting lamina area) showed no pair-wise correlation with ecological performance (Table 3), and had much more limited partial effects when PCA1 was taken into account in the path model (Fig. 7). The result of this analysis is consistent with the generalization from the literature (reviewed by Kitajima & Poorter, 2008); shade tolerance is achieved through a long leaf lifespan and a high juvenile survival rate, whereas specialization to high-light environments is achieved through faster growth. The model shown in Fig. 7 also included wood density, which is widely recognized as a key functional trait associated with shade tolerance and the life history strategy of tropical trees (Augspurger, 1984; van Gelder et al., 2006; Wright et al., 2007; Poorter et al., 2010). In the best-fit path model, wood density had a strong direct effect on sapling survival, as well as an indirect effect through its association with tissue-level leaf toughness. It is interesting that PCA1 had a negative direct effect on sapling survival (β = −0.63, P =0.02), even though its indirect positive effect through leaf lifespan (β = 0.84 × 0.40 = 0.34) and its correlation with wood density (r =0.89, P <0.001) resulted in a strong positive pair-wise correlation with sapling survival (r =0.69, P =0.001, Table 3).
Ontogenetic concordance and shifts in leaf biomechanical traits
Each leaf trait showed strong positive interspecific correlations between seedling, sapling and adult stages (Table 4). The only exception was leaf size, which typically became larger from seedling to sapling, but to various degrees amongst species. Seedlings of small-seeded species were constrained from having large leaves because large leaves cannot be built from small seed reserves. The leaves of larger saplings may be associated with longer internodes (Ackerly & Donoghue, 1998; Poorter & Rozendaal, 2008) that are expected to contribute to rapid height growth, especially in treefall gaps (Kitajima & Poorter, 2008). However, the negative direct effect of PCA3 (representing lamina area) on sapling growth rates suggests that more studies are necessary for a better understanding of the ecological significance of leaf size. The overall concordance of leaf toughness at both tissue and structural levels means that the leaf toughness of adults is a good predictor of the leaf traits and ecological performance of juveniles. This ontogenetic concordance may reflect a strong natural selection on leaf toughness at the juvenile stage, as has been found for other leaf traits (Poorter, 2007).
The ontogenetic concordance stemmed from the common patterns of trait value shifts, as summarized in Fig. 6. From seedlings to saplings, tissue density and tissue-level toughness did not change significantly. The increase in LamT, therefore, was apparently the solution to increase the structure-level toughness in order to support larger sapling leaves. The sapling-to-adult transition exhibited strikingly different patterns compared with the seedling-to-sapling transition (Fig. 6). In most species, LamT decreased from saplings to adults; yet LMA and Ws increased because the increase in tissue density and toughness overcompensated for the decrease in LamT (Table 4). A meta-analysis by Thomas & Winner (2002) also found that LMA typically increased from saplings to adult trees, although it is not clear whether changes in LamT or lamina density were responsible in the studies included in their analysis. Gras et al. (2005) found that, in four Eucalyptus species, LamT increased without a change in tissue-level toughness from juveniles to adults, which differs from the pattern summarized in Fig. 6. The decrease in LamT observed in the current study could be partly explained by a reduction in leaf size in some species whose saplings had large leaves (e.g. Stylogene, Erythrochiton, Aspidosperma, Urera and Heliocarpus). However, leaf size did not decrease in all species. Leaf traits, including tissue density and mechanical strength, can exhibit phenotypic plasticity in response to a light environment (Terashima et al., 2001; Onoda et al., 2008), and species differ in the ontogenetic trajectories of light availability measured as CE (Poorter et al., 2005). Greater vapour pressure deficit and wind in the upper canopy than in the understory may also affect the optimal leaf morphology. These observations suggest that light and other microenvironmental factors need to be documented in order to better understand the mechanistic reasons underlying the ontogenetic shifts of leaf mechanical traits.
A hierarchical approach that distinguishes between tissue-level traits and size-related traits, as well as matter distribution from mechanical strength, can elucidate the mechanistic links between leaf traits and their functional significance. Structural equation modeling supported the path models that depicted multivariate causal relationships among traits contributing to leaf toughness, as well as their links to ecological performance (Figs 1, 7). Tissue properties, including tissue density and fracture toughness of laminas and veins, were central to the structure-level leaf toughness linked to antiherbivore defense, leaf lifespan, sapling survival and growth rates, and regeneration light requirements of tropical tree species. LamT was a significant contributor to lamina work-to-shear, but was unrelated to ecological performance and regeneration strategies of the species. The leaf traits that contributed to leaf toughness showed ontogenetic shifts that changed directions and magnitudes between seedling-to-sapling and sapling-to-adult transitions (Fig. 6). Yet, the species’ rank order remained such that tissue-level toughness and the density of adult leaves predicted the regeneration ecological characteristics of juveniles. These results suggest the importance of distinguishing between tissue-level and structure-level toughness; otherwise, one may draw the erroneous conclusion that thickness per se has a defensive role, partly because thickness is often viewed as an essential aspect of sclerophyllous leaves.
We acknowledge Frans Bongers, Danaë Rozendaal and Diego Adamo for help with collection of part of the leaf data, Peter Lucas for generous support with fracture toughness measurement techniques, and Yusuke Onoda, Jared Westbrook, David Ackerly, Peter Grubb and an anonymous reviewer for constructive discussions and comments on the manuscript. LP was supported by a full fellowship and KK by a visitor’s fellowship from the Wageningen Graduate School Production Ecology and Resource Conservation.