1A height–light trajectory (HLT, a fitted curve relating canopy exposure to tree height) was determined for populations of individuals of each of 47 tree species in a Liberian lowland rainforest. The HLTs were compared and related to tree allometry and adult stature. Crown exposure was measured for 7460 trees and related to tree height using a multinomial regression analysis. Individual trees were followed for 2.8–9.8 years.
2The trajectories of the 47 species were compared with the average vertical light profile in the forest canopy. Evidence was found for the existence of all nine trajectories hypothesized on the basis of three possible light environments (high, intermediate and low) for juveniles and adults. The classical paradigm of pioneer vs. shade tolerant, based on seed and seedling responses, does not therefore apply to post-seedling stages.
3The majority of the species followed the vertical light profile in the forest canopy, starting in low light environments in the juvenile stage and ending up in high light environments in the adult stage. Only two species complied with the classic notion of whole-life shade tolerants and whole-life shade intolerants (one each).
4The predictable vertical light gradient in the forest canopy has led to a close association between adult height, light trajectories and allometric traits. Large-stature species tend to have relatively slender stems and narrow crowns, and therefore realize a faster temporal and height-related increase in crown exposure.
5Tree species have different height–light trajectories when they grow from seedling to adulthood. This may have profound repercussions for our current views on plasticity and adaptation, light partitioning and species coexistence, and on silviculture and management.
Forest succession, composition and dynamics are often explained based on the light requirements of tree species for establishment, survival and growth (Finegan 1984; Pacala et al. 1996). Traditionally, foresters in both temperate and tropical zones have classified tree species into groups according to their shade tolerance in the regeneration phase, leading to the classical dichotomy of pioneer vs. shade-tolerant species (Swaine & Whitmore 1988). Strong early environmental requirements in combination with high mortality rates may determine when, where and under what conditions seedlings of tree species are found. However, species may differ in their tolerance of shade at successive stages in their life cycle (Oldeman & van Dijk 1991; Grubb 1996; Bazzaz 1998). The ‘regeneration niche’ of a species, as defined by Grubb (1977), embraces all stages in the regeneration process, and up till now there have been very few quantitative studies of the requirements of tree species beyond the juvenile stage (e.g. Clark & Clark 1992).
While every tree may have its own unique light trajectory from seed to adult tree, a general classification can be used to indicate the possible alternative pathways during a tree's life (individual ontogenetic trajectories). In Fig. 1 we simplify these possibilities by considering four critical life-history stages and three light levels. After dispersal, a seed may be found under low, intermediate or high light conditions. The seedling growing out of this seed may be found in lower, similar or higher light conditions than the seed, as may be the juvenile and adult stage of that individual, depending on the changes in the individual tree's light environment. Trajectories can be described in the same way at the population level (population or species height–light trajectory, HLT), where the light levels experienced by seeds, seedlings, juveniles and adults may differ, depending on the light requirements for germination and establishment, and on the light dependency of survival, growth and reproduction. Species thus may show different shifts in their light requirements in different stages of their life cycle. The most extreme trajectories are formed by the outer pathways in Fig. 1: on the left, vertical arrows represent species that grow up and mature in the low light conditions of the forest understorey (traditionally referred to as strict shade-tolerant species), whereas species that establish and grow up in the bright light environment of gaps (light demanding or pioneer species) are on the far right. In between there is a whole gamut of potential light trajectories: 81 hypothetical pathways result from only four discrete life-history stages and three light levels (Fig. 1) and, in reality, the two axes are continuous, leading to an infinite number of possible pathways.
In tropical rain forest there is a strong and predictable vertical gradient in light availability: a progressive exponential extinction of light occurs when incoming radiation from above is intercepted by successive leaf layers in the forest canopy. Trees therefore should encounter brighter light conditions as they increase in height. The null hypothesis for a height–light trajectory thus would be to change from shade when small to bright light when tall. What processes lead to possible deviations of species HLTs from this null model? Species may either deviate from the average light environment right from the beginning of their life cycle, or such differences may only become apparent in later stages. Experimental evidence shows that both pioneer and non-pioneer species may germinate under a wide range of environmental conditions (e.g. Kyereh et al. 1999; Peña-Claros 2001; Pearson et al. 2002), contradicting the classical ideas (cf. Swaine & Whitmore 1988). Selection occurs therefore mainly in later stages of the life cycle, when the species are growing towards the canopy. From a physiological point of view, trees need brighter light conditions when growing in size: the ratio between photosynthesizing and respiring tissue declines with plant size, leading to an enhanced whole-plant light compensation point (Givnish 1988). The height-dependent changes in the whole-plant light compensation point may differ amongst species, and the species may follow therefore different height–light trajectories.
Despite its central role as a working model in forest ecology, little quantitative information is available on the shade tolerance of tree species. Species are often subjectively classified in two or three groups, based on the ecological knowledge and best-educated guesses of the researchers. The last decade has seen a growing body of publications, in which the light environments of tropical tree species have been quantified using PAR sensors, hemispherical photographs, or qualitative estimates of the light environment (e.g. Welden et al. 1991; Clark & Clark 1992; Clark et al. 1993; Lieberman et al. 1995; Davies et al. 1998; Rose 2000; Poorter & Arets 2003). Most studies, however, focus only on one life-history stage, or consider only few species (but see Hawthorne 1995). A community-wide approach is needed to provide the necessary resolution for revealing ecologically and statistically sound patterns, and to gain insight into the extent to which tree species partition the light gradient. Lieberman et al. (1995) took such a community-wide perspective, and considered trees over a large size range. They found that 14% of the species indeed occurred in darker or brighter conditions than expected, whereas 86% had a random distribution with respect to light. However, by the nature of their analysis, they excluded a priori the possibility that species may switch light requirements with tree height (cf. Fig. 1).
If such height-related shifts exist, they may have profound repercussions for our current views on plasticity and adaptation (Reich 2000), light partitioning and species coexistence (Hubbell et al. 1999), and silviculture and management (Lamprecht 1990). Species that experience large changes in light requirements with height might be expected to have greater trait plasticity than species that consistently occur under high or low light conditions (Popma et al. 1992; Grubb 1998; Reich 2000). Similarly, the influence of light partitioning might seem weak when only the seedling stage is considered, but might be stronger when crossovers in light requirements occur as species increase in size (Sack & Grubb 2001). Finally, silvicultural interventions such as liberation thinning might be more refined if we know for what species it is needed, and at what ontogenetic stage it should be applied.
Here we draw on a large data set from a Liberian rain forest, in which height and crown exposure were measured for 7460 trees. Temporal changes in crown exposure were followed for a 6-year period. We describe the height–light trajectories of 47 species using multinomial logistic regression analysis. Monitored trees had a diameter at breast height of 10 cm or over, and our analysis is therefore confined to the juvenile and adult stage (Fig. 1).
We address two questions. First, do species differ in their height–light trajectories from the average vertical light profile in the forest, and if so, how common is each of the hypothetical pathways identified? Secondly, are differences in these trajectories related to the maximal adult stature and allometry of the species?
Materials and methods
study site and measurements
The field study was carried out in the lowland wet evergreen forest (cf. Bongers et al. 2004) of Grebo National Forest in south-east Liberia (5°30′ N, 7°30′ W). Total annual rainfall is about 2000 mm, with a dry season (< 100 mm rainfall month−1) from December to February. Average canopy height is 35 m, the average stem density is 385 stems ha−1, and species richness is 58 species ha−1 (for trees > 10 cm d.b.h.) (Poker 1993).
Twenty 1-ha permanent sample plots (PSPs) were randomly established in an area of 900 ha, between 1978 and 1985. The nearest neighbour distance amongst plots averages 440 m (range 210–990 m). Fifteen to 20 years before plot establishment, selective logging (one tree ha−1) occurred in six plots. The other plots show no signs of recent human disturbance (Poker 1993). All trees with a diameter at breast height (d.b.h.) larger than 10 cm were measured for their diameter, tagged and identified. The height to the top of each tree was measured with a telescope pole for heights up to 15 m, and with a clinometer for heights > 15 m. Crown width was determined as the average diameter of the east–west and north–south cross-sections of the crown. The crown exposure was classified on a five-point scale (Dawkins & Field 1978): 1 = no direct light, 2 = lateral light, 3 = partial overhead light, 4 = more than 90% of the crown area receives direct overhead light, 5 = emergent crown with direct light from all directions. Estimates of this crown exposure index (CE) are repeatable and accurate (Clark & Clark 1992) and there is a good correlation between CE and canopy openness (Davies et al. 1998) or the amount of incident radiation as estimated with hemispherical photographs (Clark et al. 1993). The survival, diameter and CE of the trees were re-measured after 2.8–9.8 years (mean 6.5 years).
The height–light trajectories were analysed with a multinomial logistic regression (cf. Sheil, Chave, Salim, Vanclay and Hawthorne, personal communication). Such a regression classifies subjects based on values of a set of predictor variables (Norušis 1997) and is more general than logistic regression, because the dependent variable can include more than two categories. This analysis allows us to predict CE at first census as a function of tree height. The probability of a tree (pi) being in a certain CE class i (i = 1, 2, 3, 4) is a function of its height fi(h):
The probability of CE class 5 can be calculated as:
p5 = 1 −p1−p2−p3−p4
Several functions can be used to relate CE to the height of the tree. We used both a simple polynomial function (fi(h) = ai + bi × height), which allows for a monotonic increase of CE with height, and a quadratic polynomial function (fi(h) = ai + bi × height + ci × height2), which allows for local minima and maxima. Only the 47 species that had a sufficient number of individuals (n≥ 30) were included. The quadratic polynomial function gave a significantly better fit than the simple polynomial function for five out of 47 species. The goodness of fit is indicated by the Nagelkerke r2, which indicates what proportion of the variation is explained by the regression model (Nagelkerke 1991). The predicted mean CE at any given height was subsequently calculated as the weighted average of the five CE classes:
CE = p1 × 1 + p2 × 2 + p3 × 3 + p4 × 4 + p5 × 5
For each species we calculated this predicted mean CE for 5-m height intervals between 5 and 50 m height.
The 20 plots differ in their average crown exposure (one-way anova, F = 36, P < 0.001). A difference in HLT among species could arise if species were to occur only in one or a few plots with extreme light levels. We used an index of dispersion to evaluate whether species had a clumped distribution with respect to plots. For each species the index of dispersion was calculated as the ratio of the variance to the mean of the stem density in the different plots (cf. Fowler et al. 1998). Most of the species had a random or regular distribution over the plots, but five had a significantly clumped distribution (χ2 > 33.5, P < 0.05). Of these, Stachyothyrsus stapfiana occurred only in one plot, and Calpocalyx aubrevillei occurred in six plots and the HLT is therefore potentially biased. The other three species (Craterispermum caudatum, Diospyros sanza-minika and Gilbertiodendron preussii) occur in more than 10 plots and plot-specific effects on HLTs are likely to average out.
Species differ substantially in the speed with which they gain access to light by realizing an increase in height. The height-dependent increase in CE was calculated as the slope of the predicted CE against height (in units CE per metre). We calculated the temporal change in CE (in units CE per year) as the difference in CE between second and first measurement, divided by the monitoring period. Differences in temporal change in CE among species were tested with an anova.
Average crown exposure and height-dependent and temporal changes thereof were related to the maximal tree height (Hmax) and to stem slenderness and crown diameter of the species. The Hmax was calculated as the 95th percentile of the height values in the species’ population, thus correcting for outliers. This is important, because it is easy to overestimate the height of tall trees. Stem slenderness and crown diameter were calculated, respectively, as the tree height and crown diameter attained by an individual with a d.b.h. of 15 cm (Poorter et al. 2003).
An average vertical light profile of the forest was described using the crown exposure of all 7460 trees that were encountered in the PSPs. A 95% confidence envelope of this light profile was created for height class intervals of 5 m (2.5–7.5 m, 7.5–12.5 m, etc.). For each height class interval the crown exposures of 17 trees (the average number of trees per species per height class interval) were randomly drawn from the data set and averaged. This procedure was repeated 100 times. For each class interval we determined the 95% confidence interval (based on the 2.5 and 97.5 percentiles) for the average crown exposure of trees in that class interval.
We evaluated whether the HLTs of the species differed from the average vertical light profile in the forest, by comparing the CE at the juvenile and adult stage with the corresponding confidence intervals. Average crown exposure of a 5 m tall tree was used to represent juvenile light requirements. For eight species (Table 1) the smallest individuals were between 7.5 and 14 m tall. For these species the juvenile crown exposure was extrapolated, based on the regression equations. As adult light requirements we took the average crown exposure of trees that had attained their maximal height. For each life stage we compared whether the species had a crown exposure that was above, within or below the 95% confidence interval for trees of the same size class in the average profile.
Table 1. Crown exposure (CE) characteristics of 47 rain forest tree species. The species are ordered based on their maximal tree height. Adult stature (Hmax), juvenile and adult crown exposure, average changes in CE over the height range interval (CE change height-related) and over the monitoring period (CE change temporal) and N are given. The Hmax is based on the 95th percentile of the population. Crown exposure is determined with a multinomial logistic regression analysis for juvenile trees (5 m tall) and adult trees (average exposure at maximal tree height). For eight species (indicated with asterisks) the smallest individuals were between 7.5 and 14 m tall. For these species the juvenile crown exposure has been extrapolated, based on the regression equations. Letters following the CE indicate whether species have a relatively high (H), similar (Intermediate, I) or low (L) CE compared with other trees of the same height class interval. The Nagelkerke r2 indicates how much of the variation is explained by the multinomial logistic regression model, and is given for the significant regression lines. Note that insignificant regressions indicate that, on average, the CE is relatively constant over the height interval. Two species are only known by their vernacular name. Nomenclature follows Jongkind (2004)
Height related (m−1)
Cola lateritia K. Schum.
Craterispermum caudatum Hutchinson
Cola buntingii E.G. Baker f.
Baphia bancoensis Aubrév.
Syzygium gardneri Thw.
Garcinia afzelii Engl.
Rothmannia munsae (Schweinf. ex Hiern) Petit
Strephonema pseudocola A. Chev.
Cola nitida (Vent.) Schott & Endl.
Diospyros mannii Hiern.
Memecylon lateriflorum (G. Don.) Bremek.
Placodiscus boya Aubrév. & Pellegr.
Enantia polycarpa (DC) Engl. & Diels
Diospyros gabonensis Gürke
Carapa procera DC
Calpocalyx brevibracteatus Harms
Stachyothyrsus stapfiana (A. Chev.) J. Léonard & Voorh.
All statistical analyses were carried out using SPSS 10.
crown exposure at the stand, species and individual level
For the whole forest stand, the crown exposure increases in a sigmoid fashion with the height of the trees (Fig. 2a). Trees that are 5 m tall have an average crown exposure of 1.4, and 50-m trees an average of 4. On average, trees reach the canopy (i.e. they have a crown exposure of 3) when they are 30 m tall. The largest increases in crown exposure occur between 20 and 40 m height.
Individual canopy species, such as Gilbertiodendron preussii, may follow a similar type of trajectory with a sigmoid increase in crown exposure (Fig. 2b), but understorey species, such as Cola buntingii, remain mostly in low light conditions and show a slow increase in crown exposure (Fig. 2c). Although there are clear average height–light trajectories, there is considerable scatter: thus many species can be found at all possible light levels, and species may occur in a wide range of light environments at any given height (as shown for G. preussii in Fig. 2b).
light trajectories and strategies
We assigned the species to different light strategies, based on their average crown exposure as juveniles and adults, and on the switches they make during this transition between low, intermediate and high light compared with the average light profile in the forest canopy (Figs 3 and 4). Each of the nine possible light trajectories proposed in the upper part of Fig. 1 was found in at least one species (Fig. 3). The most common trajectories follow a continuous or sigmoid increase in crown exposure with height, and often reflect the overall community trajectory (Fig. 2d).
Most species (79%) have an intermediate crown exposure at the juvenile stage (Table 1, Fig. 4) with the remainder equally split between low and high relative CE. At the adult stage most species have a relatively low CE (57%). Only one species has a relatively low crown exposure as both juvenile and adult (Gilbertiodendron preussii, Fig. 3a), and this can be regarded as a whole-life shade-tolerant species. One species has a relatively high crown exposure as juvenile and adult (Aphanocalyx microphyllus, Fig. 3i), and this can be regarded as a whole-life shade-intolerant species. Many species (23%) have intermediate crown exposure as juveniles and adults (Fig. 3e), and simply follow the vertical light gradient in the forest canopy. Nearly half of the species (49%) start at intermediate levels and end up in relatively low light conditions as adults (Fig. 3d). Three species have relatively high CE as a juvenile, and a relatively low CE as an adult (Fig. 3g), although, in absolute terms, this means that they remain in constant light levels. Only one species (Syzygium gardneri) shows a significant absolute decrease in crown exposure with an increase in height (Fig. 3d).
Species ranking in crown exposure of juveniles is to a large extent maintained when trees increase in height. Crown exposure is strongly correlated when 10 m and 20 m trees are compared (r = 0.68, P < 0.001, n = 43) (Fig. 5a), with CE20 being consistently higher that CE10 (delta CE = 0.42, paired t-test, t = 8.26, P < 0.001, d.f. = 42). This correlation becomes weaker and disappears when larger size classes are considered (e.g. for 10 and 30 m tall trees r = 0.31, P = 0.09, n = 32; and for 10 and 40 m tall trees r = −0.16, P = 0.61, n = 12). Between the juvenile stage and the adult stage species crossovers in CE occurred in 564 out of 1081 possible pairings (52%, Fig. 5b). The species with the most extreme pairwise crossovers in CE are Anthonotha fragrans and Calpocalyx aubrevillei, which both have much greater CEs for adults than for juveniles (Table 1).
height-related changes in ce and its underlying mechanisms
Species-specific slopes of CE against tree height varied from −0.5 to 0.9 units CE per 10-m increase in height (Table 1). Species also differed significantly (−0.01 to 0.12 units CE per year, Table 1) in temporal changes over the monitoring period (anova, F46,6093 = 4.78, P < 0.001, r2 = 0.36), with Stachyothyrsus stapfiana showing an extremely large increase. As all its individuals were found in a single 1-ha plot, and this plot experienced significantly higher changes in CE than the other 1-ha plots, we excluded this species from further analysis with respect to temporal changes. Although temporal and ontogenetic changes in CE were not related to each other using a parametric correlation (rp = 0.22, P = 0.14, n = 46), a non-parametric correlation showed significance (rs = 0.30, P = 0.047, n = 46).
Tree allometry appears to play a key role in getting access to higher light levels. Species with slender stems or narrow crowns realize larger changes in CE, both height related (Fig. 6a,b) and temporally (Fig. 6c,d).
Crown exposure is positively related to Hmax over a large part of the vertical gradient (CE10 m height vs. Hmax, r = 0.36, P < 0.05, n = 47; CE35 m height vs. Hmax, r = 0.68, P < 0.001, n = 21, Fig. 7a). At larger heights the relationship between crown exposure and Hmax disappears, probably because of the small differences in crown exposure, and small sample size (n ≤ 12). Adult stature is closely linked to allometry, and height-related changes and temporal changes in CE. Large-stature species have more slender stems (r = 0.74, P < 0.001, n = 42, Fig. 7b), a larger height-related increase in CE (rs = 0.50, P < 0.001, n = 47) and a larger temporal increase in CE (rs = 0.29, P < 0.05, n = 46), as expected.
We found evidence for all nine hypothesized light trajectories, indicating that the pioneer–shade tolerant dichotomy is far too simplistic, and that it is worthwhile to look beyond the tolerances of seedlings and saplings.
crown exposure at the stand, species and individual level
At the stand level, the crown exposure increased in a sigmoid fashion with the height of the trees. The steepest changes in crown exposure occurred between 20 and 40 m, indicating the range where the light gain may be large. Unfortunately however, little is known about height growth dynamics of intermediate-sized trees. Sterck & Bongers (1998) found that intermediate-sized trees of 20 m height take large risks to get access to the canopy. They stretch out towards the canopy, and have height-diameter ratios that are close to the ratios under which they would buckle under their own mass.
Individual trees of a species occur under a wide range of light environments (Fig. 2b, cf. Lieberman et al. 1995; Bongers & Sterck 1998; Davies et al. 1998; Poorter & Arets 2003). This has led, for example, Lieberman et al. (1995) to conclude that ‘Instead of high specificity and narrow tolerance we find broad ecological amplitude; and rather than niche separation, we find a pattern of extensive overlap among the great majority of the species in the assemblage’. The question is whether high specificity and narrow tolerance should be expected from long-lived, slow-responding organisms such as trees growing in the dynamic environment as found in tropical rain forests. Mortality rates among large trees are low, and the senescence process is slow (Zagt & Werger 1998). Trees may take several years before they show morphological responses to the light environment in which they are growing (Sterck et al. 1999), and light-demanding species may therefore persist for several years in a low light environment before they die. At the same time canopy gaps are formed at a regular rate. Gaps at the forest floor are formed at a rate of 1% of the ground area per year (cf. Jans et al. 1993), whereas gaps at 20 m height are formed at a rate of 12% per year (Hubbell & Foster 1986). Over time, individual seedlings and juveniles of shade-tolerant species, especially long-lived ones, therefore have a high chance of growing, at least for a while, in a canopy gap. What counts from an evolutionary point of view, though, is whether species differ in average population-level light environments and we found that there were indeed striking differences in height–light trajectories (Fig. 3).
light trajectories and strategies
All nine of the light trajectories proposed for the juvenile–adult transition (Fig. 1) were found in at least one of our 47 species (Fig. 4), but only two strategies were common: species that follow the forest light levels (and thus the null model), and species that shift from following the forest light levels when small to relatively low levels when tall. The latter group contains typical subcanopy species that get stuck under other, taller, canopy trees. Surprisingly few species complied with the classic notion of whole-life shade tolerants (2% occurred consistently in low light) and whole-life shade intolerants (2% consistently in high light) (Figs 3 and 4). Extreme switches are rare although three species started at relatively high light levels and ended up at relatively low ones, compared with the forest average (Fig. 5g). The relative changes in Nesogordonia papaverifera and Uapaca guineensis were particularly marked but, in absolute terms, their crown exposure remained the same as height increased. Taller trees have a higher respiration load, and we might therefore expect that that these species would approach their whole-plant light compensation point, and eventually die.
In absolute light terms only one species (Syzygium gardneri) made a switch from high to low light (Fig. 3d). Species that switch from high to low light between the seedling and juvenile stage have been described as ‘gamblers’ by Oldeman & van Dijk (1991), and as ‘cryptic pioneers’ by Hawthorne (1995). They may germinate, establish and mature in the high light conditions of gaps, before becoming overtopped by faster growing pioneer species, but may then be able to persist for a long time in the shade (Dalling et al. 2001). Examples of such species are Microdesmis puberula and Myrianthus liberica in the Ivory Coast, Piper amalago in Mexico (Oldeman & van Dijk 1991), and Alseis blackiana in Panama (Dalling et al. 2001). Actually, many more species may perceive a temporal decline in crown exposure over time, either because they establish in gaps and are overgrown in the building phase (Clark & Clark 1992; Sterck et al. 1999), or because they grow towards the canopy, get stuck in the shade of an overtopping crown, and have to wait till this tree eventually dies off (Fig. 3g,i).
Species ranking in (absolute) crown exposure is to a large extent maintained with height increase, probably because most species follow the height-related null model, especially over short height ranges (Fig. 3). Over a larger height range, however, this consistency in ranking gradually disappears. This indicates that for some species the deviations from the null model accumulate with size. When species approach the canopy, individual light levels converge to similar values. We conclude that, when species are compared at two different height levels, rank reversals in irradiance are not common but do occur, especially among large-stature canopy species, but when species are compared at two different life-history stages (juveniles and adults) rank-reversals are commonplace (52% of the cases).
height-related changes in ce and its underlying mechanisms
Twenty-four per cent of the individuals experienced a change in absolute CE over the monitoring period. At the population level, species show large temporal shifts in average crown exposure. At first sight, the average change in CE (−0.01 to 0.12 units CE year−1, Table 1) might seem trivial, but if these rates are maintained they may have important consequences for light partitioning at reasonably short time-scales. If all species start with a random distribution with respect to light, and an average CE of 2, then the fastest declining species will end up after 20 years with an average CE of 1.8, whereas the fastest increasing species will end up with an average CE of 3.4. With species life spans that exceed 100 years for most species such differences may result in large crown exposure differences, especially when growth rates are high. According to our expectation, temporal shifts in CE were positively correlated with the height-dependent change in CE of the species. Species that show a steep height-dependent increase in CE also realize rapid temporal increases in CE. In contrast to our expectations, however, temporal shifts in CE were not related to the light requirements of the species (Pearson's r between temporal changes in CE and CE5m = −0.04, P = 0.79, n = 47). High-light species did not have a larger temporal shift in CE than low-light species, probably because high-light species were already concentrated in high-light environments by the juvenile stage (Fig. 3g–i).
Species-specific differences in height–light trajectories might be brought about by differences in low-light mortality, height growth and allometry. Low-light mortality has only a limited influence on the average light trajectories of juvenile and adult trees (Poorter et al., unpublished data). This is in sharp contrast to the seedling stage, where mortality rates provide an important life-history filter, determining under what light conditions species are found (cf. Li et al. 1996; Davies 2001; Peña-Claros 2001; Montgomery & Chazdon 2002).
Unfortunately, we do not have data on height growth rates of the trees. The importance of rapid vertical expansion is underscored by the fact that species with slender stems and narrow crowns realize both a faster temporal and a faster height-related increase in crown exposure (Fig. 6). By having little diameter growth and limited lateral crown expansion, these species might be able to invest more biomass in effective height extension, and take advantage of a steep vertical light gradient. For these species the extension function of tree architecture (an efficient height increase per unit biomass invested) is more important than the light interception function (Poorter et al. 2003). They gamble on reaping the future benefits (achieving a higher crown exposure and attaining reproductive size), instead of trying to enhance current light interception (King 1990; Kohyama & Hotta 1990; Sterck et al. 2001; Kohyama et al. 2003).
Adult stature partly determines the light conditions to which a species is eventually exposed. For instance, tall species, by definition, grow to positions with high light levels, in most cases having started from low light when small. We expected therefore that the nine strategies would differ in their adult stature and found that that indeed was the case (anova, F8,38 = 4.2, P < 0.001). We also found that large-stature species had higher population-level CE (r = 0.93, P < 0.001, n = 47), and that CE difference between juveniles and adults increased with Hmax (Fig. 7a). Hmax is a simple indicator of several population-level, allometric and biomechanical characteristics and reflects differences in stem slenderness as well as in CE levels. This underscores the fact that Hmax is an important life-history trait for which tree species differentiate (Thomas 1996b; Thomas & Bazzaz 1999; Turner 2001; Westoby et al. 2002). Adapting stem allometry is an effective strategy for exploiting the steep, vertical light gradient in the forest. Given this predictable and stable gradient, it is perhaps not surprising that evolutionary responses have led to a close association between Hmax, light trajectories and allometric traits.
We did not consider seed, seedling or sapling stages, which are important phases in the life cycle of the tree, where many changes occur. Many well-known commercial species, such as the Entandrophragma and Khaya species, may establish in the understorey and grow to 1 m height, but need a gap to recruit successfully to larger size classes (Hawthorne 1995; Siepel et al. 2004). We expect the inclusion of these stages to allow therefore for the occurrence of more complex light trajectories, and additional strategies (cf. Fig. 1). These diverse strategies may help species coexistence in a tropical forest environment where light availability is a critical factor for survival and growth.
We are thankful to J. Poker and the GTZ for allowing us to use the Grebo data collected by Hannsjörg Wöll and Liberian colleagues, and to David Burslem, Peter Grubb, Lawren Sack and an anonymous reviewer for their enlightening comments. Douglas Sheil, Jerome Chave, Agus Salim, Jerry Vanclay and William Hawthorne kindly allowed us to use their methodology to analyse ontogenetic light trajectories. LP was supported by Veni grant 863.02.007 from the Netherlands Organization of Scientific Research (NWO).