Differences in growth trajectory and strategy of two sympatric congeneric species in an Indonesian floodplain forest ¹

Study species

The genus Pterospermum Scherb. is characterized by medium to large trees and consists of about 20 species distributed from India to southern China and the Moluccas (Kochummen, 1972). According to Kochummen, these species are fast-growing, light-demanding trees that regenerate only in canopy gaps and on the fringes of forests. However, in our study site in the Indonesian tropical floodplain forest, we found many Pterospermum trees under closed canopies. In particular, we found two species (P. javanicum and P. diversifolium) growing under these conditions. These species form part of a continuous canopy at about 30 m aboveground and can become part of the emergent layer.

Study site

We performed our study in an equatorial floodplain forest in Labanan (Berau, East Kalimantan, Indonesia; latitude 1°52′30″ N, longitude 117°12′00″ E). The forest was designated as a 100 ha forest reserve in the Labanan concession of PT. Inhutni I, Indonesia. Records of precipitation at the Tanjun Redeb airport (ca. 20 km from the study site) from 1971 to 2000 give an average annual precipitation of 2012 mm.

A 100 × 100 m permanent sample plot was established within the forest in 2002 at an altitude of 30 m, and all trees greater than 4.8 cm DBH (diameter at breast height, 130 cm above the ground) were tagged with an aluminum number plate; the trees were then identified, their DBH was measured, and their location was mapped (E. Suzuki, Faculty of Science, Kagoshima University, unpublished data). Endertia spectabilis Steenis & de Wit of the Fabaceae (Caesalpinioideae) was abundant within the plot. Dipterocarp trees from the genera Shorea, Dryobalanops, Parashorea, and Anisoptera were also present. Numerous species belonging to the Euphorbiaceae, Ebenaceae, Annonaceae, Sapindaceae, and many other families formed the main continuous canopy at a height of about 30 m. P. javanicum was the fifth most abundant tree species in the plot; P. diversifolium was the 34th most abundant species. We found 97 and 265 individuals of P. diversifolium and P. javanicum, respectively, in the 1-ha plot. The values of the index of skewness for the frequency distribution of tree heights in the plot were positive for each species, which suggests that some big trees were present with many small trees, generating the inverse J-shaped frequency distribution typically observed for shade-tolerant species with plentiful regeneration.

Nondestructive sampling

All Pterospermum trees in the plot were identified to the species level, then were tagged using numbered plastic tapes. Each tree's location was mapped, and tree height was measured in August 2002. Tree height was defined as the height of the top of the crown for trees greater than 4.8 cm in DBH. For smaller trees, the same value was defined as stem height. In addition, the diameters at ground level and the crown widths of all Pterospermum trees that appeared to be sound and that showed no evidence of past breakage were measured by averaging two measurements, and the height at the lowest leaf (or branch, if present) was recorded. We also recorded the presence or absence of branches. To increase our sample of large trees, we added 11 trees that lay outside the plot and measured tree height, height at the lowest leaf (or branch), the stem diameter at ground level and crown diameter (again, using two measurements). The stem diameter (D, in centimeters) and crown width (C_r, in meters) were defined as the geometric mean of two measurements of the respective diameters. The crown depth (C_d, in meters) was defined as the difference between tree height (H, in meters) and the height of the lowest leaf (or branch).

Destructive sampling

Thirty-five and 28 trees of P. javanicum and P. diversifolium, respectively, with stem heights ranging from 20 to 150 cm, were destructively sampled to permit a detailed analysis of allometric relationships for the juvenile trees. These samples were collected from around the plot under closed canopies. Only trees that appeared sound were selected for sampling. All sampled P. diversifolium lacked branches (hereafter, these are referred to as “monoaxial”), but 89% of the sampled P. javanicum had branches. A crown projection photograph (top view) of all samples was taken before sampling. The juvenile trees were then cut at ground level, and the following parameters were measured: stem height, stem diameter at ground level and crown diameter (two measurements each, as described in the previous section), height at the lowest leaf and branch (whichever one occurred first), number of branches originating at the main stem, number of leaves on the tree, and length (from petiole base to leaf tip) of the longest leaf on the tree. The leaf area and the dry masses of petioles, leaf blades, stems, and branches were also measured. Stem height and the length of the longest leaf on the tree were measured for an additional 14 trees for P. diversifolium with heights ranging from 200 to 600 cm. To determine the leaf size of the mature trees, we measured the lengths of 10 leaves from a mature tree taller than 30 m for each species. From these data, we determined stem height, stem diameter, crown width, crown depth, oven-dry stem mass (M_s, in grams), leaf mass (M_l, in grams), support mass (M_sup = petiole dry mass + branch dry mass, in grams), and total mass (M = M_s + M_l + M_sup, in grams) for each sample tree.

Data analysis

We compared the allometric relationships for the two species for development of the entire plant using data derived from nondestructive sampling. To examine interspecific differences in juvenile allometric relationships in more detail, we used data derived from destructive sampling. For our analysis of the juvenile allometries, we examined three combinations of allometric pairs: (1) length dimensions (the top three pairs in Table 2), (2) length and mass dimensions (the fourth and fifth pairs in Table 2), and (3) mass dimensions (the sixth through eighth pairs in Table 2). For the analysis of whole-plant development, we lacked mass data and thus examined only allometries relating to length dimensions.

We first tested the linearity of the allometric relationship. The linearity model for an allometric relationship between the sizes of two parts (X and Y) of an organism used a simple model of the form:

where A and b are parameters obtained by a Model II linear regression (reduced major-axis regression, RMA) of the natural log-transformed values of X and Y (LaBarbera, 1989; Niklas, 1994). The nonlinearity model for an allometric relationship used a generalized allometry (Aiba and Kohyama, 1996, 1997; Thomas, 1996):

where A and b are parameters that approach those of the simple allometry for small values of Y, and Y* is a parameter that represents the asymptotic maximum value of Y. The parameters are obtained by applying an iterative, nonlinear regression program that minimizes the sum of squares of the residuals (RSS) using values of Y transformed using natural logarithms. In this case, the use of a nonlinear function necessitates the use of a Model I technique because there is no generally accepted form of Model II for nonlinear regressions (Thomas, 1996). The expanded allometry has fewer degrees of freedom because of more constants, and the RSS in Eq. 2 is smaller than that in the simple allometry (Eq. 1).

We examined the linearity and nonlinearity of the allometric relationships by means of an ordinal extension of the analysis of variance (Snedecor and Cochran, 1980):

with degrees of freedom equal to (1, n − 3). When nonlinearity was statistically supported by the F test (P < 0.05), we adopted the expanded allometry for our analysis. Although it is debatable whether the F test can be applied to nonlinear regressions, the approach seems justified in our study because this approach generated nearly identical results to those generated by means of an F test using sequential polynomial regressions (Aiba and Kohyama, 1996; Thomas, 1996).

Because the F test showed that all the juvenile allometries were linear or provided a better fit by means of simple allometry than by using expanded allometry, we used the simple allometry for our interspecific comparison. For this comparison, we applied an ordinal ANCOVA test, which is commonly used to compare allometric relationships (Kohyama, 1987; Kohyama and Hotta, 1990; Kohyama and Grubb, 1994). Although this test is based on the Model I regression method, we chose to use it for two reasons: (1) we were not interested in specifying the values of the regression parameters, but rather in detecting differences in parameters between the species, and (2) the related species would presumably show similar errors or deviations.

The differences in the slope (b) of the allometric line were tested first to detect homogeneity of slopes. If the slopes did not differ, the difference in the adjusted Y intercept of the allometric line was tested using the same slope for both species. Species with a greater slope show a greater increase in Y per increment of X, and species with a greater Y intercept (for the same slope) have a greater Y for any particular value of X.

Because the F test showed that all allometric relationships for whole-plant development, with the exception of the D–C_r relationship for P. diversifolium, were nonlinear or provided a better fit by means of the expanded allometry, we compared allometric relations for the two species using the expanded allometries. For this comparison we used an ordinal extension of the analysis of variance:

with degrees of freedom equal to (3, Σn − 3).

Evaluation of self-shading

We adopted a simplified approach to empirical measurement of the degree of self-shading (DSS) caused by upper leaves within a crown (Chazdon, 1985; Yamada et al., 2000a). The DSS value (expressed as a percentage) is defined as the ratio of the sum of overhead projections of the leaf area of all leaves within a crown to the sum of projected areas of leaves overlapped (shaded) by upper leaves; a crown with no leaves shaded by upper leaves would have a DSS value of 0%. This ratio is based on the simple assumption that a given leaf undergoes no self-shading if no upper leaves overlap it horizontally (i.e., if their vertical projection does not overlap that of the lower leaf).

We calculated DSS values for each of the destructively sampled trees. These samples had grown under closed canopies, and under these conditions light coming from angles nearer to the horizon is likely to be intercepted by the stems and crowns of neighboring trees. Most of the light captured by a tree's leaves, therefore, would come from angles nearer to the vertical, and reducing the extent of self-shading could significantly increase the tree's whole-crown carbon gain.

Leaf margins of all leaves in photographs of the crowns taken from directly overhead were traced onto tracing paper, and the leaves drawn on the tracing paper were painted. Leaf margins of shaded parts were likewise traced and painted. Leaves completely shaded by upper leaves and not visible in these views were accounted for by photographing the crowns with the shading leaves removed. All these papers were scanned, and the total area and total shaded leaf area for all leaves within a crown were determined by counting painted pixels using an area meter programmed in Photoshop 5.0 (Adobe Inc., San Jose, California, USA).

Interspecific differences in whole-plant development trajectories

The tree size at which the first branches appeared differed between the species, with P. javanicum initiating branching at lower heights than P. diversifolium (Fig. 1). Some trees of P. diversifolium had not yet developed branches at heights of 6–7 m, but all trees of P. javanicum taller than 2 m had branches.

For both species, the taller monoaxial trees had the larger leaves, but after the initiation of branching, leaf sizes decreased with tree height (Fig. 2). The ANCOVA in which species was treated as a factor, leaf size was the variate, and H was the covariate yielded a nonsignificant interaction term and a significant species term (F = 22.65, df = (1, 53), P < 0.001); the leaf size of P. diversifolium was larger than that of P. javanicum of the same height. Besides this, the size at which branching began was larger in P. diversifolium than in P. javanicum. Consequently, the maximum leaf size of the former was also larger than that of the latter: it reached about 80 cm. The lengths of 10 leaves from a mature tree averaged 19.71 cm, with a SD of 2.69 cm for P. diversifolium vs. 11.77 cm (SD = 2.17 cm) for P. javanicum. Monoaxial P. diversifolium increased its total number of leaves slowly as tree height increased, whereas P. javanicum rapidly increased the number of leaves as a function of tree height after the initiation of branching (Fig. 3). Therefore, monoaxial trees deployed relatively few large leaves on the main stem, whereas branched trees produced relatively many small leaves on branches. Hence, the monoaxial trees expanded their crown by producing larger leaves, whereas branched trees expanded their crowns by the elaboration of branches.

The curves for diameter vs. height (D–H) differed significantly between the species (F = 6.66, df = (3, 276), P < 0.001). The curve for P. javanicum had a smaller slope and a greater asymptotic (maximum) value for H than in the curve for P. diversifolium (Table 1, Fig. 4A). There was no significant difference in the diameter vs. crown depth (D–C_d) curves between the species (Table 1, Fig. 4B). The diameter vs. crown width (D–C_r) curves for P. diversifolium and P. javanicum differed significantly (F = 3.09, df = (3, 276), P < 0.05); the curve for P. javanicum had a greater slope and a smaller asymptotic (maximum) value of C_r than did the curve for P. diversifolium (Table 1, Fig. 4C).

Interspecific differences in juvenile allometries

The simple allometric relationship best described the juvenile allometries. The slopes of the regression curves for the D–H relationships differed between the species; the curve for P. diversifolium had a significantly greater slope than the curve for P. javanicum (Table 2, Fig. 5A). The D–C_d and D–C_r relationships for P. diversifolium and P. javanicum did not differ significantly (Table 2, Fig. 5B, 5C). However, the D–C_r curve for P. javanicum had a greater (less negative) Y intercept than the curve for P. diversifolium, and this difference was nearly significant (P = 0.057).

The slopes of the total mass vs. height (M–H) relationships differed significantly between the species; the curve for P. diversifolium had a significantly greater slope than that for P. javanicum (Table 2, Fig. 6A). The intercepts of the M–C_r relationships differed significantly between the species; P. javanicum had a greater (less negative) Y intercept than P. diversifolium (Table 2, Fig. 6B). In the curves of the M–M_s relationships, the intercepts of the two species differed significantly; P. javanicum had a greater (less negative) Y intercept than P. diversifolium (Table 2, Fig. 7A). In the curves for the total mass vs. leaf mass and vs. support mass relations (M–M_l and M–M_sup, respectively), the slopes differed significantly; the M–M_l curve for P. diversifolium had a steeper slope than in the curve for P. javanicum, whereas the opposite was true for the M–M_sup curves (Table 2, Fig. 7B and 7C).

Leaf display

Monoaxial P. diversifolium had relatively high DSS values, especially for saplings with H taller than 0.8 m, the DSS values were mostly greater than 10% (Fig. 8). A few monoaxial P. javanicum had comparably high values of DSS, but the DSS values for branched P. javanicum were lower than 10% in most cases. The median DSS values for P. diversifolium and P. javanicum were 12.6 and 7.19%, respectively, and these values differed significantly (Mann-Whitney U test, P < 0.05).

Growth strategy of juveniles

The most rapid height growth of stems can be attained by producing a slender stem, with a high degree of taper from the base of the stem to its apex, as well as by reducing the amount of energy allocated to horizontal growth (i.e., branches), maintaining a low specific stem mass, or a combination of these approaches (Poorter and Werger, 1999). The monoaxial growth habit is considered to be an adaptation that favors rapid height growth because it reduces the amount of energy allocated to horizontal growth (Givnish, 1978). Because a large part of the strength of the petiole and rachis derives from turgor pressure and fibrous materials, development of these structures consumes less energy than the production of woody branches bearing the same load. Therefore, a monoaxial growth habit with large, expendable leaves (Givnish, 1984) maximizes the amount of energy available for vertical growth at a given stem height and thus promotes rapid height growth. Consequently, juvenile P. diversifolium that retained a monoaxial growth habit for longer (until a larger tree size) than P. javanicum and deployed large leaves directly on the stem during monoaxial growth, follow a dynamic function of tree architecture to enhance the opportunistic use of light after a disturbance rather than a static function that optimizes leaf display under current lighting conditions.

The observed D–H and M–H relationships for juveniles support the hypothesis that P. diversifolium's growth strategy emphasizes height growth. Curves for this species have a significantly higher slope for the D–H and M–H relationships than P. javanicum. This means that P. diversifolium attains more height growth than P. javanicum for a given increment in D and M. The slopes of the D–H curves for P. diversifolium and P. javanicum were 1.90 and 1.33, respectively. These values are much greater than the values for models based on constant stress or elastic similarity, which are 0.50 and 0.67, respectively (McMahon, 1973; Sposito and Santos, 2001). This means that the safety margin for stem rigidity (i.e., the ratio of actual stem diameter to the theoretical calculated minimal diameter to prevent elastic buckling) decreased with increasing tree size. Similar trends have been reported for other tropical Sterculiaceae (Yamada et al., 2000b, 2001) and for many other tree taxa (King, 1990; Sposito and Santos, 2001). Rapid height growth would be favored over stem stability by juvenile trees in tropical forests, in which vertical gradients in relative light availability are steeper than in forests outside the tropics (Yoda, 1974).

Juvenile P. diversifolium had smaller crown widths and higher values of DSS than P. javanicum. Because monoaxial growth must initially expand crown width by displaying larger leaves, horizontal growth is limited compared with trees that adopt a branching growth habit. Furthermore, the monoaxial growth habit develops large leaves toward the stem's leader, leading to significant self-shading of lower leaves. Consequently, the static function of tree architecture of P. diversifolium is inferior to that of P. javanicum. Unlike P. diversifolium, P. javanicum branched well starting at a younger age. The branching growth habit expands crown area by making and expanding branches with small leaves, thereby distributing the leaves more widely to achieve a low DSS and create a monolayer crown shape. Therefore, P. javanicum's architecture is more effective in terms of increasing survival at a given height or under darker conditions than is the case for P. diversifolium.

Growth trajectories

Many researchers have suggested that a nonlinear function for the generalized allometry would better fit observed D–H relationships for whole-plant development (Aiba and Kohyama, 1996; Thomas, 1996). The maximum heights we estimated using a generalized allometry for D–H relationships were lower for P. diversifolium than for P. javanicum. This is because the architecture of juvenile P. diversifolium is suitable for rapid initial height growth, but height growth ceases when these trees reach the canopy layer and the trees rarely pass through the layer to become emergents. King (1996) found that long-lived species showed a greater increase in crown diameter when they reached the upper end of their height range. Yamada et al. (2001) observed a switch from stem elongation to radial growth with increasing tree size in another tropical species in the Sterculiaceae (Scaphium longiflorum Ridley); small trees elongated rapidly but thickened less during rapid stem elongation, whereas large trees elongated relatively little but thickened rapidly to strengthen the stem's mechanical stability. The relative importance of rapid height growth and stem stability for a tree would depend on tree size. An alternative explanation of the nonlinearity of the D–H relationship involves the relationship's dependence on reproductive maturity (Thomas and Ickes, 1995; Thomas, 1996). The D–H relationship, leaf proliferation, and leaf size are all known to show pronounced allometric discontinuities associated with the onset of reproduction (Alvarez-Buylla and Martinez-Ramos, 1992; Thomas and Ickes, 1995; Thomas, 1996). More studies are needed to assess the allometric discontinuities related with reproductive onset and to identify the size at which reproductive onset occurs for various species.

Pterospermum javanicum continued height growth after reaching the continuous canopy layer and passed through this layer to become emergent. In contrast with D–H relationships, P. javanicum ceased crown expansion after reaching the canopy layer, although juveniles allocated considerable energy to expanding their crown area. The asymptotic maximum crown width for P. javanicum, estimated using a generalized allometry, was smaller than that of P. diversifolium. Thus, the allometric strategies adopted by juveniles appear to be largely independent of those adopted by larger trees. This suggests that diverse pathways to reaching maturity should exist among tropical trees, a finding that would not have been revealed had we analyzed only juvenile architectures and allometries. The study of whole-plant development trajectories is strongly recommended for researchers interested in clarifying the diverse growth strategies that may be adopted by tropical trees passing from the seedling stage to maturity.

Coexistence mechanisms for two congeneric species within a community

Differentiation of growth strategies between P. diversifolium and P. javanicum, assessed by means of tree architecture, can be expected to lead to equilibrium coexistence in a community if the two species differ in their regeneration niche as a function of understory light conditions. This is because each phenotype potentially provides competitive advantages, and these advantages depend on understory light conditions. In the competition for growth and survival between these two species, P. diversifolium is hypothesized to outperform P. javanicum in bright microsites because its architecture is better adapted to rapid height growth. In contrast, P. javanicum is hypothesized to outperform P. diversifolium in shaded microsites because its architecture is better adapted to light assimilation under those conditions. Therefore, we can expect that the spatial distribution of P. diversifolium in a forest community will be restricted to brighter microsites, whereas P. javanicum will be more widely distributed where understory light is limited. Furthermore, stem elongation by P. diversifolium in bright microsites is likely to be more rapid than that of P. javanicum. To verify these hypotheses, further comparative studies of spatial distribution, survival, and growth in relation to understory light conditions will be needed.