Differences in the dry-mass cost of sapling vertical growth among 56 woody species co-occurring in a Bornean tropical rain forest

study site and species

The study was conducted at Lambir Hills National Park, Sarawak, Malaysia (4°2′ N, 113°50′ E; 150 m a.s.l). Consistently high temperatures, with no distinct dry season, characterize the climate of this region. The mean annual temperature and annual precipitation were 26 °C and 2700 mm, respectively. The 6949-ha park was covered by primary lowland mixed dipterocarp forest on nutrient-poor sandy clay or clay-rich soils. Dipterocarpaceae dominated the canopy and emergent layers; Euphorbiaceae, Burseraceae and Myristicaceae dominated the lower layers.

Over a 2-week census period in December 2005, we collected three saplings (1·5 ± 0·2 m tall) of as many species as possible growing under closed canopy. In this site, the light environment is relatively homogeneous as far as the canopy is closed (c. 6% of the light in open sites; Aiba & Nakashizuka 2005; also see Delissio & Primack 2003). Individuals suffering severe herbivory or with evidence of regrowth after stem breakage were rejected. Species with arching or multiple stems were excluded because they would have severely complicated the analysis. Thus, we collected specimens of 16 branched overstory species, 13 monoaxial simple-leaved overstory species, 15 monoaxial compound-leaved overstory species, three branched understory species, six monoaxial simple-leaved understory species and three monoaxial compound-leaved understory species. All of the branched saplings were simple-leaved because we found none with compound leaves within the specified size range.

We defined understory species as those < 15 m in maximum height, whereas overstory species were ≥ 15 m in height at maturity. Maximum heights were obtained from published records (Whitmore 1972; Soepadmo & Wong 1995). For a few species with no published information on maximum height, we used the typical adult height at the study site. For details of the species, see supplementary data.

measurement of sapling structural traits

After measuring the height, crown width in two directions at 90° to one another (including the greatest width), and diameter at ground height in two directions at 90° to one another, we harvested the above-ground plant material. The saplings were then divided into leaves, support tissues (branches, petioles and rachises) and trunk. All plant material were weighed (after being oven-dried at 60 °C for at least 7 days) to obtain trunk dry mass (M_T), leaf dry mass (M_L) and support tissue dry mass (M_S). The total above-ground dry mass (M_W) was the sum of M_T, M_L and M_S, and crown mass (M_C) was the sum of M_L and M_S. Some typical sized and fully expanded, but not senescent, leaves (2–26, depending on the leaf size) were dried separately after digitally scanned to calculate individual leaf area (A_IL) and specific leaf area (SLA). Scanned images were processed using the software package LIA32 (ver. 0·376β1, K. Yamamoto), and SLA was calculated from the area and mass. We estimated the total leaf area (A_L) of each sapling by multiplying M_L by SLA. The projected crown area (A_C) was approximated as an ellipse. The trunk diameters measured from two angles at ground level were geometrically averaged (D). Leaf area index (LAI) and wood density index (WDI) were defined as A_L/A_C and M_L/trunk volume approximated as a cone, respectively. All parameters, definitions and units are summarized in Table 1.

We considered M_W, the total current above-ground dry mass of 1·5-m tall saplings, the extension cost. A limitation of this definition was that biomass lost with leaf and branch turnover, was not included. For the saplings of 31 dipterocarp species analysed in a separate study, the lost biomass of saplings 1·3–1·5 m tall was estimated as up to 62% of the current biomass and < 20% for most species (Aiba, unpublished data). This value is relatively small compared with the eightfold variation in the M_W of the 56 species (supplementary data). The potential effects of lost biomass on our results will be discussed below in more detail.

data analysis

Because of the careful selection process, plant heights were quite similar among species (1·51 ± 0·06 m, mean ± SD; cf. supplementary data). Therefore, it was not necessary to control for differences in the mean height of samples in the following analyses. Data were log₁₀-transformed to normalize positively skewed distributions. The M_S of the only unbranched sessile-leaved species, Agrostistachys longifolia (Euphorbiaceae), was assigned a value of 0·01, although it was actually 0.

We examined the relationship between structural traits and adult stature or crown type using two-way anova of both raw data (i.e. a cross-species analysis) and phylogenetically independent contrasts (PICs; Felsenstein 1985) of the 12 structural traits. The two factors were adult stature (understory or overstory) and crown type (branched, monoaxial simple-leaved or monoaxial compound-leaved), and 12 dependent variables, i.e. structural traits, were described above. For the cross-species analysis, we calculated R² values after separating the effect of species from the residuals to evaluate the extent of intraspecific variation.

anova of PICs was performed to examine the effects of phylogenetic constraints on the relationship between sapling structure and adult stature or crown type. We constructed a phylogeny for the 56 species (supplementary data) using the Web application Phylomatic (Webb & Donoghue 2005), and calculated PICs using the Web application COMPARE (Martins 2004). To assure the bifurcation of the phylogeny at all nodes, which is required for calculation of PICs, we arbitrarily resolved polytomies by the insertion of very short branches (Martins 2004). Categorical variables, i.e. adult stature and crown type, were transformed to dummy variables before the calculation of PICs.

Type II regression (using the standardized major axis method because variables X and Y in our analyses were subject to error; Sokal & Rohlf 1995) was used to examine interspecific relationships between A_L and M_S or M_C in a comparison of dry-mass cost required to display a given leaf area for each of the three crown types. In this analysis, we did not separate understory from overstory species because no significant interactions were observed between adult stature and crown type for any of the structural traits (two-way anova; P > 0·05). Agrostistachys longifolia was excluded from this analysis because it lacked support tissue entirely and its inclusion produced highly skewed results. To compare relationships among the three crown types, we performed ancova-like analyses, followed by post hoc multiple comparisons of standardized major axis regressions. We used the (s)matr software package (version 1; Falster, Warton & Wright 2003).

Principal components analysis (PCA) was used to determine the dominant axes of structural trade-off among the 56 species. SLA, LAI and WDI were omitted from this analysis because they were secondary parameters calculated as ratios of other traits. All statistical analyses without standardized major axis regression and the calculation of PICs were performed using the software package R (version 2·01).

intergroup differences in extension cost and other structural traits

The M_W of 1·5-m tall saplings, i.e. the extension cost, varied considerably among the 56 species (77 ± 40 g; cf. supplementary data). Swintonia acuta had the lowest extension cost and weighed 34 g at 1·5 m height. In contrast, Barringtonia dolicophylla had the most costly structure and weighed 274 g at 1·5 m height.

The amount of variance explained by our model, which included the effect of adult stature, crown type and species, was > 70% of the total variance for most structural traits, showing that intraspecific structural variation is relatively small compared with interspecific variation. No significant interactions were observed between adult stature and sapling crown type in cross-species comparisons (two-way anova; Table 2). In comparisons of the three crown types, no significant differences in extension cost were detected. Monoaxial simple-leaved saplings had significantly smaller M_S than other saplings, but the difference did not result in a smaller M_C. The M_L of monoaxial simple-leaved saplings was significantly greater than that of monoaxial compound-leaved saplings. The A_IL of monoaxial saplings was significantly greater than that of branched saplings. SLA and WDI were relatively small, and LAI and D tended to be large in monoaxial simple-leaved saplings compared with at least one other crown type.

Compared with overstory species, understory species had a significantly larger M_W, M_T and D (Table 2, Fig. 1). In contrast, no significant differences in other traits were observed, including all crown characteristics.

The analysis of PICs generally agreed with the results from the cross-species analysis, although some effects that were significant in the cross-species analysis were not significant and some interactions became significant. Lineages that contained many understory species tended to have larger M_T and D through evolutionary processes. Such lineages also had larger M_W, but the effect was not significant. Lineages that contained many branched species had smaller M_L, A_IL, LAI and WDI, and larger A_C than monoaxial lineages.

effects of crown type on dry-mass cost to display leaves

No significant differences were found among the regression slopes of the relationships between log A_L and log M_S calculated using the standardized major axis method (Table 3, Fig. 2a). The intercepts (after accepting the common slope) differed significantly among the three crown types, with the largest for branched saplings and the smallest for monoaxial simple-leaved saplings. When we evaluated the display cost by M_C relationships, we found a significant difference among the slopes (Table 3). The slope for monoaxial simple-leaved saplings was significantly greater than that for branched saplings. As the order of the intercepts was opposite to that of the slopes, the three regression lines crossed, resulting in relatively small differences among the three regressions (Fig. 2b).

ordination of the 56 species using their structural traits

The structural variation of the 56 species was determined using PCA. The first principal component explained 62% of the total variance and summarized a trade-off between vertical extension cost and light interception and/or mechanical resistance, showing strong negative correlations with M_W, D, A_L, M_T, M_L and A_C (Fig. 3a). The second principal component explained 14% of the total variance and was positively correlated with M_S, indicating that M_S is independent of M_W, unlike other traits. Sapling groups differing in adult stature or crown type were not separated along the first principal component (Fig. 3b). All groups showed similar variation along this axis, indicating that extension cost is independent of crown type. In contrast, the second principal component separated saplings into different crown types, indicating that the only structural difference among the crown types was M_S.

sapling crown type and extension cost

Monoaxial saplings have been considered adapted to rapid height growth at the expense of lateral expansion, which enhances the survival capability under shaded conditions (Givnish 1978; Coomes & Grubb 1998; Yamada et al. 2005). Monoaxial saplings could undoubtedly achieve low extension costs if the dry-mass cost of leaf display using petioles or rachises was considerably smaller than that of leaf display using branches, and if the monoaxial characteristic was independent of other structural traits. However, our results indicate that these requirements are not satisfied, and consequently, no significant difference was detected in extension cost between monoaxial and branched saplings (Table 2, Fig. 1). Monoaxial simple-leaved saplings met neither requirement. Although these saplings required an M_S smaller than that of branched saplings (by 4·5 g) to display an A_L of 0·269 m² (the grand mean A_L for all 56 species), this advantage diminished to 1·9 g for M_C (Type I regression, data not shown; cf. Fig. 2). This structural advantage is trivial compared with a mean M_W of 77 g, and was likely not responsible for ecologically important differences in height growth rates. This was mainly attributable to the considerable leaf thickness in monoaxial simple-leaved saplings (Table 2, Fig. 1), which may be essential for the mechanical support of large leaves (Shipley 1995; Niinemets, Portsmuth & Tobias 2006). Thus, the thickening of leaves functions as a substitute for branches in monoaxial simple-leaved species, and the two methods of support do not differ significantly in dry-mass cost. In a comparative analysis of 44 herbs and woody seedlings in an Estonian deciduous forest, Niinemets et al. (2006) also reported a convergence of the leaf-display cost as a result of a trade-off between support tissue within leaves and branches. Furthermore, the monoaxial simple-leaved saplings had larger M_T, M_L and D, all of which increased the extension cost, compared with other types of saplings, although the reason for this is not clear. Altogether, the disadvantages of a heavier trunk and thicker leaves completely outweighed the limited advantage of leaf display, making the monoaxial simple-leaved sapling structure relatively inefficient for vertical growth compared with the other sapling types (Fig. 1).

The dry-mass cost of displaying a given leaf area in monoaxial compound-leaved saplings was not significantly different from that in branched saplings (Fig. 2), and the advantage in M_S in displaying a leaf area of 0·269 m² was only 1·4 g. Thus, rachises of compound leaves do not appear to be economically ‘expendable branches’, as predicted by Givnish (1978), and they have virtually no advantage over genuine branches in displaying leaves, at least for young individuals in shade.

Therefore, at least among juveniles of the size examined here and under shaded conditions, unbranched species are not economically advantageous in extension cost compared with branched species, and not necessarily inferior in their ability to intercept light. However, it could be that monoaxial species are superior to branched species in extension cost under sunny conditions as a result of structural plasticity. Monoaxial saplings may be opportunists, simply awaiting gap formation while growing slowly, as found by Coomes & Grubb (1998) for Amazonian tree species, rather than ‘optimists’ (sensu Kohyama 1987) that grow faster, even in shade, to gain a competitive advantage following gap formation in the near future.

adult stature and sapling extension cost

The M_W, M_T and D of 1·5 m tall understory trees were significantly greater than those of taller species (Table 2, Fig. 1), supporting previous reports. All of these trends were common to the three crown types, indicating that these differences are not related to the predominance of a certain crown type in understory species (Table 2). Thicker stems can provide mechanical resistance against disturbance caused by falling debris, which often causes heavy damage to small woody plants in the forest understory (King 1990; Poorter & Werger 1999). This would be especially advantageous for understory species, which complete their life cycle beneath the canopy.

However, we failed to find any significant differences between understory and overstory species in crown architecture characteristics related to light interception. This is a consistent observation in comparisons of young saplings (Kohyama & Hotta 1990; King 1996; Poorter & Werger 1999), although understory species often have a wider crown than overstory species in larger juveniles (Aiba & Kohyama 1997; Kohyama et al. 2003; Poorter et al. 2003). Therefore, at least for young saplings, resistance against physical disturbance, rather than the ability to intercept light, seems a likely explanation for the higher extension cost of shorter species.

The omission of lost biomass from the data is a limitation of this study. In young saplings whose branch systems are simple, the ratio of lost biomass to current biomass is largely determined by the ratio of the length of naked trunk (i.e. trunk without leaves and branches) to plant height. The length of naked trunk was significantly large in understory saplings compared with overstory saplings, and in monoaxial species compared with branched species (anova, P < 0·05; data not shown). Therefore, the inclusion of lost biomass is assumed to disproportionately increase the extension cost of understory saplings and monoaxial saplings. These conditions suggest that the major findings, i.e. no advantage of monoaxial saplings in extension cost and higher extension cost of understory species, are robust to the omission of lost biomass.

trade-off between vertical growth efficiency and ability to survive under shaded conditions

We did not find a trade-off between extension cost (M_W) and crown characteristics that promote light interception (M_C, A_L and A_C) among any of the pre-assigned categories, but the trade-off was evident when all species were considered together (Fig. 3). Species of the three crown types were equally variable in this trade-off (Fig. 3), indicating that the trade-off occurs within each crown type. This supports our basic finding that the regeneration strategy is independent of crown type.

Although previous studies have emphasized the importance of the distribution of biomass as a mechanism underpinning the functional trade-off between height growth and light interception (Kohyama 1987; King 1990), it was not extensive allocation to M_L or M_B, but rather, proportional increases in M_L and M_T, that were essential for the trade-off. Extension cost (M_W) was virtually independent of the leaf mass ratio, trunk mass ratio, and support mass ratio (Pearson's correlation; R² < 0·03). Although it has not been explicitly discussed (Kohyama 1987; King 1990; Kohyama & Hotta 1990), the previously noted interspecific variation in crown size and trunk thickness may also result from interspecific variation in total dry mass, rather than variation in biomass distribution.

What are the ecological consequences of this considerable variation in total dry mass among co-occurring tree species? We can suggest two contrasting predictions. If the variation in net production per individual or its allocation to aerial parts is much smaller than that in total above-ground mass, the height growth rate will generally follow the dry-mass cost. In this case, variation in above-ground mass must be a reflection of a strategic trade-off between future competition and current survival, as previous studies have postulated. With spatiotemporal heterogeneity in light availability, such differentiation in regeneration strategy would lead to the stable coexistence of these species. Otherwise, larger net production based on the larger leaf area/mass may completely compensate for the disadvantage in high extension cost, and therefore, height growth rate may be independent of sapling structure (Cho et al. 2005). In this case, the reason for variation in extension cost could be much simpler: the variation is functionally neutral with respect to height growth rate. Further studies involving comparative analyses of both net production rate and height growth rate are necessary to test these alternatives and to fully understand the relationship between variation in extension cost and species coexistence.