Frank J. Sterck, Utrecht University, Section Plant Ecology, Postbox 80084, 3508 TB Utrecht, The Netherlands (email@example.com).
1 Monitoring of two canopy species Dicorynia guianensis and Vouacapoua americana (Caesalpiniaceae) in a tropical rain forest in French Guiana was used to investigate vegetative crown development at five organizational levels: leaf, metamer, extension unit, sympodial unit and whole crown. The effects of light availability and tree height on different traits were evaluated in trees < 25 m in height and compared with taller individuals (25–37 m). Path-analysis is used to illustrate the consequences of trait changes at multiple levels of organization for the whole crown level.
2 Tree height and canopy openness influenced crown development at each organizational level. Crowns in higher light levels had lower specific leaf area, greater leaf spacing, greater extension of all branches, and greater extension of the leader shoot. With increasing tree height, crowns had a lower specific leaf area, greater leaf area index and greater relative crown depth.
3Vouacapoua showed some responses to light not seen in Dicorynia. In particular, Vouacapoua increased meristem activity with light, but the lack of response in Dicorynia may be due to moderate light levels rather than inability to respond.
4 Low leaf-display costs at low light availability may enable trees to survive light suppression.
5 Light availability cannot explain trait changes with tree height. Alternative explanations for trait changes with tree height are discussed.
6 Several of the relationships between plant traits and tree height or canopy openness became non-linear when taller trees (25–37 m) were included. In these taller trees, vegetative growth was reduced at all organizational levels, particularly in Vouacapoua, which does not grow as tall as Dicorynia.
7 Qualitatively, plant responses to light did not differ between trees of different height, and were similar to seedling and sapling data in the literature. Responses were, however, quantitatively different, suggesting that small saplings cannot serve as model organisms for crown development in taller trees.
Our approach to the process in structurally complex rain forest trees defines vegetative crown development as the production, positioning and morphological properties of the various components listed in Table 1 (see also White 1979; Barthélémy 1991; Bell 1991; Room et al. 1994). These plant components characterize the different levels of organization in the crown hierarchy, and are produced and spaced according to an inherited pattern (Hallé & Oldeman 1970; Halléet al. 1978). This hierarchy remains the same as trees get taller or respond to environmental conditions, and can be used to integrate development at different organizational levels and to compare crown development among trees in different light environments and in different ontogenetic phases.
Table 1. Plant components in order of increasing size (partly adapted from Bell 1991; Room et al. 1994) and representing the various organizational levels in the crown
Region of active cell division and enlargement. There is typically a meristematic zone at the apex (apical meristem) and each node (axillary meristem) along every axis. Axillary meristems become apical meristems once they become active and produce a new shoot.
An internode with axillary bud(s) at its proximal end and one or more leaves at its distal end, but without shoots resulting from growth of axillary buds.
The vegetative axis, displayed by one period of activity of a meristem (‘flush’), either of apical or axillary origin. It consists of a sequence of metamers.
The total sequence of metamers and/or extension units produced by one meristem (also referred to as module, Prévost 1967; Halléet al. 1978).
The assemblage of leaf-supporting branches, including portions of the stem that support leaves or leaf-supporting branches.
Components at each of the various organizational levels are well known to respond to light, i.e. they are located at different positions and produced at different rates (King 1994), and have different sizes or architectures (Fisher 1986; Shukla & Ramakrishnan 1986). Theoretical predictions that crown development at low light availability will favour survival if it is associated with slow growth rates (Veneklaas & Poorter 1998), if bifurcation ratios are low and crowns are flatter (Leopold 1971), if there is reduced self-shading of the leaves (Horn 1971) and if there is higher allocation of mass to leaves relative to wood (Givnish 1988) are generally supported by empirical studies of seedlings and small saplings (e.g. Canham 1988; King 1994).
The influence of ontogeny on vegetative development is less well known and, although it has been described for some rain forest trees (Halléet al. 1978; Edelin 1990), information on the effects on bifurcation ratios, crown allometry, leaf self-shading and leaf allocation remains anecdotal and ambiguous. Trees often become more exposed as they grow taller and crown development may therefore change with tree height, just as it does with increasing light availability. This would parallel physiological leaf traits where assimilation and respiration rates are reported to increase as either tree height or light increased (Kitajima 1996; Rijkers et al. 2000). Yet some trees become less exposed with increasing tree height (e.g. Clark & Clark 1992). Moreover, trees become increasingly woody and have to be maintained by relatively low light interception (or leaf) areas. We would therefore expect any change in vegetative crown development with tree height to parallel that with declining light levels. Most of the available data relates to the pioneer species that monopolize light gaps (e.g. Alvarez-Buylla & Martinez-Ramos 1992; Ackerly 1996), rather than the shade-tolerant trees that grow in variable light conditions (Strauss-Debenedetti & Bazzaz 1996).
Ontogenetic stage and light environment are likely to influence crown development simultaneously. Although the effects of both tree size and light level have been considered for leaf morphology (Niinemets & Kull 1995; Niinemets 1997; Rijkers et al. 2000), such joint investigations on crown development are rare and usually focus on smaller individuals and few organizational levels (e.g. Canham 1988). The combined effects of development at the various organizational levels for the whole crown are also poorly understood.
We explicitly distinguished between the effects of tree height and light availability on each of the plant traits in Table 2 (i.e. at the various organizational levels) to evaluate their influence on crown growth, leaf display and leaf display costs. We studied individuals of two shade-tolerant species in French Guiana, ranging in size from small saplings to canopy trees, to test two hypotheses: (i) plant traits will vary with light availability such that trees will have slower crown growth, wider crowns, less self-shading, smaller leaf display cost, smaller leaf spacing and larger specific leaf area at lower light levels; and (ii) because an increase in tree height is accompanied by a gradual increase in light availability, trait changes with increasing tree height (up to 25 m) will parallel those seen under conditions of increasing light.
Table 2. Plant traits, ordered by organizational level. Metamer is equivalent to extension unit in Vouacapoua
Relative crown depth
Crown depth/mean crown width
Total branch extension
Total branch length produced per year
Total leader length produced per year
Ground projection of the crown area
Total leaf area
Mean leaf area × total number of leaves
Leaf area index
Total leaf area/crown area
Number of meristems
Number of apical meristems
Sympodial unit level
Number of metamers produced per sympodial unit per year
Leader metamer production
Number of leader metamers produced per year
Sympodial unit survival
Percentage of sympodial units surviving per year
Sympodial unit production
Percentage of sympodial units that is produced in preceding year
Mean length of metamer
Leader metamer length
Mean length of leader metamer
Number of leaves produced per metamer per year
Number of fallen leaves per metamer per year
Mean length between leaves
Relative leaf spacing
Leaf spacing/leaf area
Mean surface area of individual leaf
Specific leaf area
Leaf area/leaf mass
Materials and methods
Site and species
Field work was carried out in a pristine lowland tropical rain forest at the biological field station ‘les Nouragues’ (4°05′ N, 52°40′ W), French Guiana. Annual rainfall is approximately 3000 mm, with dry seasons (< 100 mm month−1) in September and October, and sometimes in March. The forest is dominated by species of Lecythidaceae, Leguminosae, Sapotaceae, Chrysobalanaceae and Burseraceae (Sabatier & Prévost 1989). The area is covered with well-drained, clayey to sandy-clayey soils on weathered granite parent material.
Dicorynia guianensis Amshoff. and Vouacapoua americana Aubl., hereafter referred to by their generic names alone, are abundant in the canopy. Such Caesalpiniaceae or other leguminous species are (co-)dominant in many neo-tropical forests and most are characterized as shade-tolerant (e.g. Schulz 1960) and as conforming either to Troll’s model of canopy architecture or to other mixed plagiotropic models (Oldeman 1974; Halléet al. 1978; Oldeman 1989). Thus, the stem leans over to support a horizontal spray of distichously arranged leaves and the process is repeated at different positions in the crown as a result of relays sprouting from the bends. Stem and branches cannot be distinguished from one another at early stages of crown development, but one branch subsequently straightens towards the vertical and continues secondary thickening growth and thus becomes the stem, while the other branches fall or develop in a more horizontal direction. Both study species are known for their commercial timber value in French Guiana and Surinam.
Figure 1 shows how different components (leaves, metamers, extension units and sympodial units, Table 1) contribute to crown development in the study species. For Vouacapoua, metamers were not considered because they are structurally variable: those at the bottom of an extension unit support scale leaves, and the ones at the top support compound leaves. For Dicorynia, however, metamers are structurally equivalent, with each producing one compound leaf, and therefore extension units cannot always be unambiguously distinguished (but see Drénou 1994).
Field work and measurements
Populations of both species were inventoried in October 1992 as part of a long-term study of tree growth (Sterck 1997). A 12-ha plot was mapped for trees with a stem diameter ≥ 10 cm at 1.30 m above ground level (d.b.h.), and a central 1.5-ha plot was mapped for trees with a stem diameter < 10 cm, but with a height greater than 0.50 m.
For each species, we selected 20 of the inventoried individuals that were less than 4 m tall, ensuring that the full range of light levels within the site was represented. Their crown development was measured from the ground, either by bending the stem over or using natural elevations within the forest. The crowns of taller individuals could only be studied from neighbouring trees (climbed using spikes or alpinist ropes). Because only a few of the inventoried individuals were accessible in this way, we included a selection of the accessible 4–25 m trees (13 Dicorynia and 20 Vouacapoua) and all five accessible > 25 m trees (Vouacapoua, 25, 35 and 37 m, and Dicorynia, 26 and 37 m) from the adjacent area outside the plots. Measurements were made in November 1992, November 1993, and November 1994 for all individuals.
The light environment was recorded using hemispherical photographs, taken in November 1993. A Canon Ti 70 body and Canon 7.5 mm/5.6 lens was mounted in a leveller and telescopic poles (up to 6 m long) were used to position the camera just above the top of each individual. The light environment of taller individuals was also measured by climbing neighbouring trees with ropes or spikes. Black-and-white prints were scanned using a Hewlett Packard Desk Scan II and canopy openness (the percentage of open sky) was calculated using the programme Winphot 5 (Ter Steege 1996). Canopy openness is used as an estimate of light availability.
Trees were marked and drawn to scale in 1992. All branches and apical meristems of individuals < 4 m were tagged and coded, as was a selection of branches plus their apical meristems (n = 10–45) within the top of the crown of taller individuals. In later censuses (1993 and 1994), we measured size, structure and production rate of each plant component (for details see below).
Production was assessed by counting leaves, leaf scars, metamers (Dicorynia), extension units (Vouacapoua) and apical meristems (or active sympodial units) that had been produced by 1994 on top of apical meristems tagged in 1992. For Vouacapoua, the number of leaves and scars included photosynthetic but not scale leaves. Although for trees taller than 4 m detailed counts were made only for the selected apical meristems, we assessed the total number of leaves and apical meristems by climbing carefully along the length of these individuals. In each individual, the axis (linear sequence of metamers or extension units) that supported the uppermost apical meristem of the crown in 1994 was defined as the leader.
Leaf area was determined (Delta Image Analysis System, Eijkelkamp 1991) as the mean value for 20 leaves taken from random positions in the crown. Lengths were measured for all metamers (Dicorynia) or extension units (Vouacapoua) produced distal to monitored meristems. Mean metamer length and extension unit length were calculated for both the whole crown and the leader separately. Lengths of sympodial units are not reported because they may have continued length growth after the last census in November 1994. We also measured the greatest crown width, the crown width perpendicular to this, and crown depth, i.e. the vertical distance between the uppermost and the bottom leaves.
Plant traits were categorized into various organizational levels (as detailed in Table 2). At crown level we used the measurements to calculate relative crown depth, total branch extension, leader extension, crown area, total leaf area, leaf area index and number of meristems. At sympodial unit level we calculated metamer (or extension unit) production, leader metamer production, sympodial unit survival and sympodial unit production. At metamer (or extension unit, in the case of Vouacapoua) level we calculated metamer length, leader metamer length, leaf production, leaf fall, leaf spacing (a measure for the average distance between leaves) and relative leaf spacing (a measure for the leaf spacing per unit area). At leaf level we calculated leaf area and specific leaf area.
Multiple linear regression was used to assess the relative importance of tree height and canopy openness in determining plant traits. Trees taller than 25 m were excluded because their inclusion caused relationships between plant traits and tree height to become non-linear in 11 out of 36 cases. In some cases, plant trait values were transformed if this improved the linearity among variables and/or the homogeneity of variances. The following regression model was used:
Here, c is the constant (intercept), the b1, b2, b3 are the regression coefficients and e is the error term.
We used path-analysis (e.g. Wright 1934; Kingsolver & Schemske 1991) to integrate the effects of tree height and canopy openness on plant traits and to show the consequences of these plant traits for the whole crown (Fig. 2). Path-analysis shows direct (or causal) effects as path-coefficients, and relations without causal effect as Pearson product moment correlation coefficients. The analysis distinguishes direct, indirect and total effects (e.g. Sokal & Rohlf 1981). In general, path-analysis indicates the change in a criterion variable (expressed in SD units) as a result of a change of 1 SD in each of its predictors, both direct and indirect. Correlation among predictors (collinearity) inflates both the path-coefficients and their confidence intervals. Consequently, the effect of a predictor on a criterion variable is either overestimated with respect to its magnitude, or neglected because it is harder to get a coefficient that differs significantly from 0 (Petraitis et al. 1996). In our analysis, we had to deal with collinearity between tree height and canopy openness, as well as among plant traits. We did not evaluate the magnitudes of significant effects since the coefficient values cannot be accurately estimated. We realise that by only distinguishing between effects that are significant and those that are not, we may erroneously have rejected actual effects (increased type II errors).
Tree height vs. canopy openness
Canopy openness of individuals < 25 m tall ranged from 0.8% to 30% (Fig. 3). For the five taller individuals it ranged from 12% to 80%. Tree height and canopy openness were correlated (Pearson’s product moment coefficient of 0.54 in Dicorynia and 0.74 in Vouacapoua, P < 0.001). Although canopy openness tended to be higher above taller trees, there was a wide range of canopy openness values for any particular tree height. In our multiple regression analyses and path-analyses we concentrated on how crown development is affected by canopy openness independent of tree height, and by tree height independent of canopy openness.
Plant trait changes with tree height and canopy openness
Multiple regression analysis was used to evaluate changes in plant traits with tree height and canopy openness (Table 3) for trees < 25 m high. For clarity, the plant trait changes are described for each organizational level.
Table 3. Regression of plant traits (see Table 2) on tree height (m), canopy openness (%), and their interaction for trees (0.5–25 m high) of Dicoryniaguianensis (n = 33) and Vouacapouaamericana (n = 40). Standardized regression coefficients are presented. When interaction effects were not significant, regressions were calculated with only height and openness. In the case of Vouacapoua, metamers are equivalent to extension units (see methods section). *** P < 0.001, ** P < 0.01, * P < 0.05, NS = model not significant. /= coefficient not significant
Data log transformed.
Not relevant for Dicorynia as leaf production is 1 by definition.
Metamer length is synonymous with leaf spacing in the case of Dicorynia.
Crown level: In both species, total branch extension increased with both tree height and canopy openness, and leader extension increased with canopy openness independent of tree height. All other traits in Dicorynia increased with height alone, whereas in the case of Vouacapoua they increased with both tree height and canopy openness, except for crown area (height only) and leaf area index (interaction only).
Sympodial unit level: Extension unit production decreased with tree height in Vouacapoua. Sympodial unit production decreased with tree height in Dicorynia and increased with canopy openness in Vouacapoua.
Metamer level (only for Dicorynia): Metamer length (and thus also leaf spacing) and relative leaf spacing increased with canopy openness independent of tree height.
Extension unit level (only for Vouacapoua): Extension unit length, leaf spacing and relative leaf spacing increased with canopy openness independent of tree height. Leaf production and the number of leaves increased with tree height.
Leaf level: In Dicorynia, specific leaf area decreased with tree height and canopy openness. In Vouacapoua, the decrease in specific leaf area required increases with both tree height and canopy openness, as illustrated by the highly significant interaction term.
Plant traits integrated over organizational levels
The path-analysis shows how tree height and canopy openness influence plant traits at a higher level through their effects on plant traits at lower organizational levels (Figs 4–7). Path-analysis does not account for interactions between tree height and canopy openness, nor for non-linear relationships, and is therefore restricted to trees shorter than 25 m.
Total branch extension depends on the number of meristems, metamer (or extension unit) production per meristem, and metamer (or extension unit) length (Fig. 4). In both species tree height contributed to the increase in total branch extension because of the greater number of meristems. In Vouacapoua, extension unit production was also affected (in this case negatively) by tree height, but the effect on total branch extension was minor, compared to the strong positive effect of the number of meristems (values of effects not shown). Canopy openness increased total branch extension in Vouacapoua due to both increased numbers of meristems and greater extension unit length. In turn, the greater number of meristems resulted from a greater sympodial unit production (Table 3). In Dicorynia, however, the increase in total branch extension with canopy openness resulted only from the positive effect on metamer length. Thus, only one organizational level contributed to the more rapid crown growth of Dicorynia with both greater tree height and with greater canopy openness, whereas two levels were involved in Vouacapoua.
Leader extension depends on leader metamer (or extension unit) production, and on the length of these plant components (Fig. 5). The increase in leader extension in Dicorynia with greater canopy openness was mainly due to increases in metamer length. Vouacapoua showed the same tendency, although the effect of canopy openness on leader extension unit length was not significant (P = 0.07).
Total leaf area depends on the number of meristems, metamer (or extension unit) production per meristem, leaf production and leaf fall per metamer (or extension unit), and leaf area (Fig. 6). The increase in total leaf area with tree height resulted primarily from an increasing number of meristems. Although extension unit production in Vouacapoua decreased with tree height, its effect was minor compared to the strong positive effect of the number of meristems. Only in Vouacapoua did total leaf area increase with greater canopy openness and this was also due to an increasing number of meristems. Ultimately, leaf area index was strongly influenced by total leaf area rather than by crown area (although this had a minor negative effect in Vouacapoua).
At the metamer (or extension unit) level, we analysed how leaf spacing (equal to metamer length in Dicorynia) and relative leaf spacing depended on other plant traits (Fig. 7). Both measures were governed by canopy openness (see also Table 3) via increased unit length. These patterns indicate shorter woody support structures per unit leaf area at lower canopy openness, and thus lower costs for producing a unit of leaf area.
Trees ≥ 25 m in height
When trees taller than 25 m were included, regressions of several plant traits on tree height were non-linear (Fig. 8). We were unable to distinguish between the effects of tree height and canopy openness on crown development in these tall trees, and can only compare their development with shorter (and generally less exposed) trees. At the crown level, non-linear relationships became apparent generally for both species between tree height and relative crown depth, total leaf area, number of meristems, and leader extension (Fig. 8a–d). There were no non-linear relationships at the sympodial unit level, neither were there for Dicorynia at the metamer level, whereas extension unit length, leaf production and leaf spacing in Vouacapoua showed maximal values around 25 m (Fig. 8e–g). At the leaf level, leaf area decreased above 25 m, particularly in Vouacapoua (Fig. 8h).
Total crown growth and leader growth (which is correlated with height growth) increased with canopy openness, independently of height in the latter. Only one or two organizational levels contributed to these crown growth responses, in contrast with another study of these species (Sterck 1999) where both measures of crown growth of juveniles (4–18 m) were higher in gaps than in the understorey because of changes at all possible levels. It may be that some traits respond only to the high light levels characteristic of gaps and these are rarely experienced by trees shorter than 25 m. Alternatively, gradual changes may become significant only when well-separated light levels are compared.
The influence of light availability on leader extension has been shown for seedlings (e.g. Popma & Bongers 1988), saplings (e.g. King 1991) and taller juveniles (Sterck 1999). The increase in height growth with size in eight Costa Rican tree species (Clark & Clark 1992) might also be due to light availability rather than size itself. Leader growth therefore generally seems to be controlled by light in trees up to at least 25 m, whereas total branch extension (which has rarely been studied explicitly) is more affected by tree height (Table 3).
Total leaf area and leaf area index increased with tree height in Dicorynia. In Vouacapoua, total leaf area increased with both tree height and canopy openness, but leaf area index was affected only by their interaction. An increasing number of meristems explained most of each effect.
Small individuals (< 10 m) of the pioneer tree Cecropia obtusifolia (Alvarez-Buylla & Martinez-Ramos 1992) show a similar increase in total leaf area with height but as a result of a more than 100-fold increase in leaf area, rather than meristem activation. Dicorynia and Vouacapoua typically have c. 20 apical meristems by the time they reach 10 m in height, and larger individuals continued to increase total leaf area by producing more apical meristems, as is the case with Cecropia individuals > 10 m in height. Similar changes may therefore be the result of different plant traits in each species.
The early branching of Dicorynia and Vouacapoua follows Troll’s model, while Cecropia persists for longer as a single pole axis and its subsequent branching pattern is described by Rauh’s model (Alvarez-Buylla & Martinez-Ramos 1992). Below 10 m, Cecropia (still in pole phase) may have a higher leaf area index than the branched Dicorynia and Vouacapoua, but self-shading is reduced by the positioning (Ashton 1978; Orians 1982) and by wider spacing of its leaves (King 1994).
The higher sympodial unit production in Vouacapoua with greater canopy openness indicates both an increased branching rate and a higher bifurcation ratio (an index of the degree of branching from one branching order to the next) with greater canopy openness (Canham 1988) and was associated with narrower crowns. Leopold (1971) and Whitney (1976) argued that such high bifurcation ratios are more efficient (in terms of stem tissue used for leaf display) for narrower crowns in higher light environments, while low bifurcation ratios are more efficient for planar crowns (which have less leaf self-shading) at lower light levels. Similar increases in bifurcation ratios at higher light availability have also been found in other studies (Steingraeber et al. 1979; Shukla & Ramakrishnan 1986).
Although juveniles of both species had larger total leaf area, higher leaf area index, and longer and relatively narrower crowns in gaps than in the darker understorey (Sterck 1999), the lack of such effects in Dicorynia in the present study may (as with crown growth) reflect the moderate light levels. Similar differences in results were found between studies of small saplings of the temperate species Acer saccharum, which showed a response only when fully exposed sites (highest light levels) were included (Steingraeber et al. 1979; Canham 1988; Bonser & Aarssen 1994).
Factors related to leaf display costs
Leaf display costs can be defined as the amount of carbon invested to produce one unit of leaf area, and depend on the relative amounts, densities and construction costs of leaf and wood material. Such costs must be reduced at low light levels if light interception is to increase, but only the size components show significant plasticity (e.g. King 1991, 1994). Leaf spacing, relative leaf spacing and specific leaf area reflect the (relative) amounts of wood and leaves along new shoots, although the impacts of secondary thickening and root growth are not considered.
Specific leaf area (a leaf area trait) is negatively related to both tree height and light availability in Dicorynia, but only to their interaction in Vouacapoua, similar to patterns observed for other species (Jurik 1986; Oberbauer & Strain 1986; Poorter et al. 1995; Rijkers et al. 2000). Everything else being equal, a higher specific leaf area at lower light levels and/or tree height will result in a greater light interception area per leaf mass, and thus contribute to reduced leaf display costs. However, morphological information alone may be misleading because many other leaf-level factors, including content of expensive metabolites such as chlorophyll, may vary (Boardman 1977; Pearcy 1987; Ellsworth & Reich 1993) and contribute to reducing costs in higher, more open canopies.
Leaf area was unaffected by either factor, but leaf spacing and relative leaf spacing both decreased when canopy openness is less; this decrease was due to shorter metamers (Dicorynia) or extension units (Vouacapoua). Leaf spacing also decreased at lower light in other tropical trees (King 1991, 1993, 1994), but there is as far as we know no comparable data on relative leaf spacing. All three traits considered appear to contribute to a reduction in the leaf display costs at lower light availability, which may enable trees to lower their whole-plant compensation point and thus survive light suppression.
From understorey to canopy
Regressions of plant traits against tree height showed linear relationships in trees under 25 m in height, while several regressions became non-linear if larger trees were also included. This apparent switch in crown development might be triggered at a certain height, greater exposure or some other environmental factor. On attaining the canopy, branching patterns may change drastically (King & Maindonald 1999) and even multiple tree-like structures (trunk-like branches, each with its own subcrown) may develop in a process referred to as reiteration or metamorphosis (Halléet al. 1978; Edelin 1990); however, no data are available on associated quantitative changes in crown development traits.
Many crown-level traits in both species lost their linear increase with tree heights above 25 m. The changes in relative crown depth and leader extension suggest that crowns which have been narrowing with increasing height then widen above 25 m, as seen for taller trees in other tropical forests (Bongers & Sterck 1998; Bongers et al. 1988; King 1996). This is probably due to the higher light levels in the upper open canopy because the change occurs at the height where light levels increased sharply (Koop & Sterck 1994) and juvenile plantation-grown trees of Dicorynia and Vouacapoua had wider crowns than forest-grown trees of the same height (F.J. Sterck, unpublished data). In the open upper canopy, where there is little chance of being over-topped, increasing crown width may enable trees to shade and thus out-compete their smaller neighbours.
In Vouacapoua, changes with tree height and canopy openness also became non-linear above 25 m for some traits at the extension unit level, indicating a reduction in vegetative crown growth; this observation is expected because maximum height is about 35 m for this species. No such effects were seen in Dicorynia where the largest trees (25 and 37 m) were still below the 45–50 m height of full-grown adults.
At the leaf level, leaf area decreased above 25 m for Vouacapoua and this may reflect adaptations to physiological drought in the open upper canopy, but there was no effect in Dicorynia. In general, adults of canopy tree species have smaller leaves than juveniles of the same species (Thomas & Ickes, 1995).
Traits generally responded to light in the same ways as previously reported for smaller saplings. Returning to our first hypothesis, under conditions of decreasing light availability, trees were (as expected) found to be characterized by slower crown growth rates, relatively wide crowns, fewer leaf layers, smaller leaf display costs, lower sympodial unit production, smaller leaf spacing and larger specific leaf areas. The absence of some responses in Dicorynia probably reflected the general absence of light levels able to induce an effect, rather than an inherent inability to respond.
Our second hypothesis, that increasing tree height and increasing light availability influence tree characteristics in a similar way, was supported by parallel responses in some traits at the crown and the leaf level. However, even in these cases changes with height may not be due to the higher light levels: at the leaf level, effects on specific leaf area are more likely to reflect physiological drought, and effects at the whole crown level were due to the influence of height on the number of apices. At the three intermediate levels (metamer, extension unit and sympodial unit), the changes in plant traits with tree height never paralleled the patterns with light availability. The decrease in the proportion of productive (leaves) to unproductive (wood) biomass with increasing tree height may counterbalance the positive effects of increasing light availability. Overall, light availability cannot be said to have been an important selection force acting on architectural changes with tree height.
Our approach also suggests that small saplings cannot serve as model organisms for the study of crown development in trees of other sizes. While the directions of responses of saplings and larger trees to light are similar, no quantitative predictions can be made. As well as changing with tree height per se, many trait responses to light show changes in magnitude, or even in direction (for trees above 25 m in height), as the tree grows. Furthermore, responses to light at the crown level are predominantly driven at intermediate organizational levels (sympodial unit, extension unit and metamer), which rarely receive attention in small saplings. The modelling of crown development in trees therefore requires data from trees of a range of sizes. In addition, even closely related species such as those studied here (same plant family and similar general branching pattern) may differ in responses of some traits. This may have far reaching consequences for modelling forest canopy dynamics on the basis of crown development of individual trees, particularly in species-rich habitats such as tropical rain forests.
We thank L. Poorter, T. Rijkers and M. Smeenge for giving field support; D. Ackerly, C.D. Canham, P.A. Jansen, D. King, E. Leigh Jr., J. den Ouden, L. Poorter, H.H.T. Prins, S. Wijdeven and J. Wright for giving comments on earlier drafts of the manuscript; M.G. Born for drawing Fig. 1. This study was supported by grant W 85-239 of the Netherlands Foundation for the Advancement of Tropical Research (WOTRO).
Received 29 September 1999 received 25 February 2000