Growth responses of 15 rain-forest tree species to a light gradient: the relative importance of morphological and physiological traits


  • L. Poorter

    1. Department of Plant Ecology and Evolutionary Biology, Utrecht University. PO Box 80084, 3508 TB Utrecht, the Netherlands and Programa Manejo de Bosques de la Amazonía Boliviana (PROMAB), Casilla 107, Riberalta, Bolivia
    Search for more papers by this author


1. Growth of seedlings of 15 rain-forest tree species was compared under controlled conditions, at six different light levels (3, 6, 12, 25, 50 and 100% daylight).

2. Most plant variables showed strong ontogenetic changes; they were highly dependent on the biomass of the plant.

3. Growth rate was highest at intermediate light levels (25–50%) above which it declined. Most plant variables showed a curvilinear response to irradiance, with the largest changes at the lowest light levels.

4. There was a consistent ranking in growth between species; species that were fast growing in a low-light environment were also fast growing in a high-light environment.

5. At low light, interspecific variation in relative growth rate was determined mainly by differences in a morphological trait, the leaf area ratio (LAR), whereas at high light it was determined mainly by differences in a physiological trait, the net assimilation rate (NAR).

6. NAR became a stronger determinant of growth than LAR in more than 10–15% daylight. As light availability in the forest is generally much lower than this threshold level, it follows that interspecific variation in growth in a forest environment is mainly owing to variation in morphology.


Rain-forest tree species are often classified into two functional groups, based on their germination and establishment requirements (Whitmore 1996). Shade-tolerant species can germinate, grow and survive in low light, whereas light-demanding species need a high-light environment for establishment. These two groups are thought to present the extremes of a continuum of responses to light (Osunkoya et al. 1994). Yet, there is little empirical evidence on how far species are indeed graded in their shade tolerance. As a result the continuum often falls apart into a quite distinct, well-known group of a few pioneer species, and the remaining shade-tolerant species, which constitute the vast majority of the tropical-forest tree species. In a rain forest in Panama, for example, it was found that saplings of 79% of the species were shade-tolerant. They were indifferent to gaps and could regenerate both in a high- and a low-light environment (Welden et al. 1991).

In tropical rain forests generally only 1–2% of the radiation above the canopy reaches the forest floor (Chazdon 1988; Clark et al. 1996). Accordingly, growth rates of seedlings in the understorey are very low. In a Costa Rican rain forest, it was shown that during a 4 year period 20% of the saplings did not show any height growth (Clark & Clark 1992). Once the canopy is opened up, light availability is enhanced, giving rise to vigorous growth. Size differences at time of gap formation may determine to a large extent the success of an individual afterwards. The larger this head-start is, the more likely it is that an individual eventually will reach the canopy (Brown & Whitmore 1992; Boot 1996; Zagt & Werger 1998). Hence, both growth and survival are crucial for successful regeneration. In this study I focus on plant responses to a light gradient and define what plant features explain interspecific differences in seedling growth.

Plant responses to light vary with position along the light gradient. Plants at the lower end of the light gradient enhance their light interception, as light becomes a limiting resource. Shaded plants have a higher biomass allocation to leaves (leaf mass ratio; LMR) and a higher leaf area per unit leaf mass (specific leaf area; SLA), resulting in a higher leaf area per unit plant mass (leaf area ratio; LAR) (Popma & Bongers 1988; Osunkoya et al. 1994). At the same time self-shading is reduced by making wide crowns with a low leaf area index (Kohyama 1991; Sterck 1997). Furthermore, they have a higher stem length per unit stem biomass (specific stem length; SSL) thus increasing height growth in a way that facilitates escape from the low-light environment (Sasaki & Mori 1981). Plants in high light, on the other hand, are faced with high radiation loads. Thus they invest more in root mass, in a way that compensates for higher transpiration losses by water uptake. The biomass production is increased by means of a higher light-saturated photosynthetic rate (Poorter & Oberbauer 1993), biomass growth per unit leaf area (net assimilation rate; NAR) (Veenendaal et al. 1996) and a high leaf turnover (Bongers & Popma 1990).

To date, most comparative studies have evaluated seedling growth at two to three irradiance levels (but see Okali 1971; Kwesiga & Grace 1986; Veenendaal et al. 1996). The qualitative responses to light are therefore relatively well known. It is less well established how seedlings respond to a light gradient; where on the gradient they show largest phenotypic responses to light, and at what irradiance they attain maximal growth. Veenendaal et al. (1996) compared seedling growth of 15 West African tree species at five irradiance levels. They found that largest changes in biomass allocation and plant morphology occurred at the lower end of the light gradient, and that most tree species attained maximal growth rates at intermediate light conditions.

At a given irradiance, species may differ in their relative growth rate (RGR). The RGR can be decomposed into a morphological component, the leaf area ratio, and a physiological component, the net assimilation rate. In a comparative study on seedling growth of 12 Australian rain-forest tree species (Osunkoya et al. 1994) it was found that in a low-light environment (2·5% daylight) interspecific variation in RGR was associated with variation in NAR, whereas in an intermediate-light environment (10% daylight) it was associated with variation in LAR. These results suggest that different components become important at different levels of irradiance.

In this study seedling growth of 15 rain-forest tree species is compared at six different irradiance levels. The following questions are addressed; (1) how do plants respond to different irradiance levels in terms of growth, biomass allocation, morphology and architecture; (2) what determines interspecific variation in RGR and how does this vary with irradiance?

Materials and methods


Fifteen tree species were selected for the study, including one palm species (Table 1). Species selection was based on presumed difference in shade tolerance (including pioneers such as Cecropia ficifolia and shade tolerants such as Theobroma speciosum) and adult stature (Theobroma, understorey species, Cariniana micrantha, emergent species). Several species were included because of their economic importance (Bertholletia excelsa provides the Brazil nut, Euterpe precatoria is harvested for its palm heart, Swietenia macrophylla, Cedrela odorata and Amburana cearensis are highly valued timber species). Hereafter species will be referred to by their generic name only, except the two Aspidosperma species, for which the full species name will be used.

Table 1.  . Summary of the species studied: the formation of a seedling bank in the forest understory, strategy (P, pioneer; I, intermediate; S, shade tolerant), adult stature (U, understorey; S, subcanopy; C, canopy; E, emergent), growth period in the experiment (days), date of initial harvest, source of plant material (S, seed; P, transplanted seedling from forest) and commercial use. Classification of the species in pioneer, intermediate and shade-tolerant species was based on the spatial distribution of saplings over microhabitats Thumbnail image of


The experiment was carried out in the experimental garden of the Universidad Técnica del Beni, Riberalta (11 ° S, 66 ° W) in the Amazon region of Bolivia. The climate in the region is characterized by a mean annual temperature of 27 °C and a mean annual rainfall of about 1780 mm with a low rainfall period (< 100 mm month–1) lasting from May until September (Beekma, Zonta & Keijzer 1996). The vegetation in the region can be classified as a tropical lowland moist forest.


Plants were grown at six different irradiance levels. Five tunnel-shaped shade cages with a semicircular cross-section were positioned perpendicular to the daily track of the sun. In this way spatial and temporal variation in irradiance within the shade cages was reduced as much as possible. Shade tunnels had a height of 1·5 m, and an area of 3 m × 3 m. Five irradiance levels were created by covering the shade tunnels with an increasing number of layers of nuetral shade netting. Each extra layer intercepted an additional 50% of the incoming radiation, thus creating five irradiance levels of about 50, 25, 12, 6 and 3% irradiance. An additional irradiance level consisted of plants grown in the open (100% of incident irradiance). Ventilation in the shade tunnels was allowed for by leaving a 10 cm slit open between the soil surface and the shade cloth. The lowest irradiance level (3%) included in the experiment is somewhat higher than generally encountered in the understorey of tropical rain forests (1–2%). Yet, the forest in the region is a bit more open than other forests, and the lowest irradiance level used in the experiment may be a fair approximation of irradiance levels encountered in large parts of the understorey.


Owing to the limited capacity of the shade tunnels, differential seed/seedling availability of the species and low seed longevity, three groups of species were grown in three different years (Table 1). Plants were either germinated from seed or collected as wildings from the forest. Wildings were about 10 cm tall, transplanted to potting bags and positioned in the shade tunnel at 3% light. Thereafter, wildings were gradually moved to higher irradiance levels, which allowed them to acclimate gradually to the new light environment. Seeds were sown in trays in the different shade tunnels, or were directly sown in the individual potting bags. Once germinated, seedlings from the trays were transplanted to large-sized polystyrene 5 litre bags (35 cm high, 15 cm diameter). The bags were filled with a mixture of one-third forest top soil and two-thirds river sand. The forest soil was used to provide a substrate with a natural composition of macro- and micronutrients. The river sand provided a texture with adequate drainage, which allowed for daily watering of the plants and facilitated harvest of the whole root system, including fine roots. By the end of the experiment, plant leaves in all but the highest light treatment were healthy and green, suggesting that the plants were not nutrient limited. The bags were placed on wooden platforms 3 cm above the soil, thus preventing roots to grow through the drainage holes of the bags into the soil.

Twenty plants per species per treatment were randomly positioned in the shade tunnels. For Aspidosperma tambopatense the 100% treatment was lacking as insufficient plant material was available. Plants were watered every other day and re-randomized within the shade tunnel halfway through the experiment. Eight plants were used for the initial harvest (mean 7·9, range 5–10) and 10 plants (mean 10·2, range 5–12) for the final harvest. For each species and treatment an initial harvest was carried out once the seedlings had exhausted seed reserves as far as they were going to, and were acclimated to their present light environment. As rain-forest trees rely upon their seed reserves for only a short period of their life (Kitajima 1996) it was preferred to start the experiment once the attached seeds had fallen off. The only exceptions were Theobroma, which still had its cotyledons at initial harvest, and Bertholletia, which has a large seed that is incorporated into the stem. Therefore initial harvest of Bertholletia was carried out when the plants attained such size that seed mass formed a minor part of total plant biomass. Acclimation of plants to their present environment is especially important for studies where plants are grown over a large range of resource availability. When plants are moved from a low- to a high-light environment, the increased irradiance may cause wilting and bleaching of the plants. In this experiment, the mean time between germination or transplanting and the initial harvest was 15 weeks (range 2–36). The mean growth period between initial and final harvest was 23 weeks (range 12–29).

At initial harvest, stem length, leaf area and leaf, stem and root mass were determined. At final harvest crown width in two perpendicular directions and height insertion point of the lowest leaf were measured. After harvesting the plants were divided into leaves, stem plus petioles, and roots. Dependent on the size of the plants, the total leaf area was determined for all leaves or for a subsample of leaves selected along the stem. For the species harvested in 1994 photocopies of the leaves were made and analysed with an IBAS image analysis system (Kontron/Zeiss, Eching, Germany) at Utrecht University. For the other species leaf area was determined with a portable CI-202 leaf area meter (CID Inc., Vancouver, WA). Leaf thickness was measured for the same leaves used for the leaf area determination. Measurements were carried out for three leaves at two positions, using a dial thickness gauge (Mitatuyo). Care was taken to measure the leaf blade in between the leaf nerves. For most of the Cecropia leaves this was not possible, as the leaves had a dense and fine venation. Afterwards plant parts were oven dried for at least 48 h at 70 °C and weighed.

From the primary data the following variables were derived: root mass ratio (RMR; root mass/total plant mass, in g g–1), stem mass ratio (SMR; stem + petiole mass/total plant mass, in g g–1), leaf mass ratio (LMR; leaf mass/total plant mass, in g g–1), specific leaf area (SLA; leaf area/leaf mass, in m2 kg–1), leaf area ratio (LAR; leaf area/total plant mass, in m2 kg–1), leaf area root mass ratio (LARMR; total leaf area/root mass, in m2 kg–1), mean leaf size (total leaf area/total leaf number, in cm2), specific stem length (SSL; stem length/(stem + petiole dry mass), in cm g–1), crown area (π× 0·25 × average crown width2, in m2), relative crown depth {RCRD, 100 ×[(stem length – height position lowest leaf)/stem length, in %)]}, and leaf area index (LAI, total leaf area/crown area, in m2 m–2). These variables refer, respectively, to biomass allocation (RMR, SMR, LMR), leaf display (SLA, LAR), the balance between investment in light intercepting organs vs water and nutrient uptaking organs (LARMR), the efficiency of biomass investment for height gain (SSL), and crown architecture (crown area, RCRD, LAI).

Mean relative growth rate (RGR; biomass growth per unit plant biomass, in mg g–1 day–1) was calculated following Venus & Causton (1979), net assimilation rate (NAR; biomass growth per unit leaf area, in g m–2 day–1) and leaf area ratio (m2 kg–1) were calculated according to the formulas given by Hunt (1978). The formula for the calculation of NAR is only valid when leaf area and plant dry mass are linearly related (Hunt 1978). For all species-treatment combinations this condition was met; mean r2 of the linear regression of leaf area against plant mass was 0·90 (range 0·68–0·99). The instantaneous relative growth rate equals the product of NAR and LAR. If a longer time period is considered, RGR and NAR present mean values over the growth period. For the equation to hold, LAR also should be calculated as a mean value over the growth period. The variance of the RGR was calculated following Causton (1991).

The whole-plant light-compensation point was estimated for each species from a linear regression of NAR against ln-transformed percentage daylight. Only the four data points of the lowest irradiance levels (3, 6, 12, 25%) were included in this analysis.


The shade tunnels in the experiment were not replicated because of logistical constraints. However, as a result of the wide spacing and large number of light levels, the variation in light between tunnels is larger than variation within tunnels, and trends in plant responses to light can be shown. For the statistical analysis it is assumed that differences in plant responses between shade tunnels are only owing to differences in radiation and related climatic variables, and not to other confounding factors. The species were grown in three different years. Inter-annual variation in climate might have influenced the outcome of the results. In the region, there is little variation in temperature between years. Variation in rainfall is neither likely to have been a problem as plants were watered every other day. Inter-annual variation in radiation did not influence the interspecific comparisons to a large extent; if the range in interspecific growth rates is large enough, like for the group of species harvested in 1995, then the same interspecific patterns emerge as when species of all three harvest groups are pooled.

Plant responses were evaluated using a two-way ANOVA, with light treatment and species as independent variables. All dependent variables were transformed to natural logerithms before analysis. The reasons were twofold. First, an assumption of ANOVA is that effects of the independent factors are additive, and not multiplicative (Sokal & Rohlf 1995). Nevertheless, it may be assumed that plants respond to an increase in resource availability in a proportional (i.e. multiplicative) way. For example, an increase in light intensity may lead, for both small- and large-sized species, to a doubled biomass in high light compared to low-light conditions. With a two-way ANOVA on non-transformed data, one would incorrectly conclude that there is a significant interaction between species and treatment, based on the wrong assumption of additivity of the independent factors. With a ln-transformation, such an interaction would no longer be found (Poorter & Garnier 1996). An additional advantage is that variances are stabilized as well.

Plant biomass may differ between treatments and species at the end of the experiment and ln-transformed biomass was therefore included as a covariable in the analysis. In this way the effect of main factors on plant variables can be evaluated by comparing plants at a similar biomass. Interaction between species and biomass was included in the analysis, as ontogenetic trajectories may differ between species. Ln-biomass was centred around the origin by subtracting grand mean plant mass at final harvest (i.e. the mean plant biomass of all species and all treatments pooled) from the individual plant biomass values. In this way plant variables are compared at the grand mean plant mass at final harvest. The ANOVA was carried out for 14 species only, as for A. tambopatense one treatment was lacking, which otherwise would result in an incomplete statistical design. All statistical analysis were carried out using SPSS 6·0 (Norus˘is 1993)


To analyse how underlying factors contribute to interspecific variation in RGR, a growth response coefficient (GRC) was calculated (Poorter & van der Werf 1998). The GRC indicates what proportion of the interspecific variation in RGR is caused by variation in one of its components. If RGR is analysed as the product of NAR and LAR, then the sum of the GRCs of NAR and LAR should equal one. Alternatively, RGR can be defined as the product of NAR, LMR and SLA. In that case, the GRCs of these three factors should add up to one as well. The GRCs for a group of species can be obtained by calculating the regression slope of ln(NAR) or ln(LAR) against ln(RGR). The advantage of the use of GRCs over the use of a correlation coefficient, is that the former allows for the evaluation of the relative importance of the growth parameters in explaining interspecific variation in RGR. In contrast, with a correlation coefficient one merely establishes that there is a relation between a growth parameter and RGR, without indicating how much of the variation in RGR is associated with variation in this parameter. For a more extensive description of GRC, see Poorter & van der Werf (1998).

The GRCs of NAR, LAR, LMR and SLA were calculated for each light level using species mean values as data points. The growth response coefficients were plotted against irradiance, thus showing how the contribution of morphological and physiological components of interspecific variation in RGR may change with irradiance.



There was a large variation in plant size and morphology among species and irradiance levels (Table 2). The explained variation by the ANCOVA model was very high (mean 0·87, range 0·69–0·98) and most factors were highly significant. All variables were significantly influenced by plant biomass, or its interaction with species. The F-value allows one to compare the relative importance of light, species and biomass for plant morphology. Most variables related to size (plant height) or to architecture (crown area, RCRD) were strongly influenced by biomass, as indicated by the high F-values. The sign of the regression slope shows how plant variables changed with biomass. Height, crown dimensions, LMR, leaf size and leaf thickness increased with biomass, whilst SLA and SSL declined with biomass. Significant species × biomass interactions for plant variables indicated that species followed different ontogenetic trajectories. Yet, these F-values were generally much lower than F-values for other factors. Light was the most important determinant of variation in leaf-related characteristics (LMR, SLA, LAR), and characteristics related to the water balance of the plant (RMR, LARMR). In contrast, SMR, leaf thickness and LAI appeared to be more species-specific.

Table 2.  . Results of a two-way ANCOVA with light (n = 6) and species (n = 7–14) as fixed effects and ln (biomass) as a covariable. F-values, significance (P), standardized regression coefficient of the common slope of ln(biomass) (b) and coefficient of determination (r2) are shown. The covariable was entered in the analysis together with the main effects. The dependent variables were ln-transformed prior to analysis. Variables are grouped into categories related to size, allocation, morphology and architecture. For most variables 14 species were included in the analysis, except for leaf thickness and architectural variables, for which only seven species were included. F-values of factors having the largest effect on a plant variable are given in bold; NS, P > 0·05; *P < 0·05; **P < 0·01; ***P < 0·001 Thumbnail image of


In general, plants attained the highest biomass under intermediate light conditions. Biomass at final harvest was highest at 25–50% light (see Appendix for results of Student–Newman–Keuls test). Stem length was closely related to biomass (Table 2) and therefore it attained the highest value at intermediate light conditions as well (Appendix). To show functional responses to light, i.e. without being confounded by biomass differences between species and treatments, graphs are shown for corrected values at the grand mean mass of all plants at the final harvest for those variables largely influenced by biomass (Fig. 1). Plant responses are typically those found along a light gradient. Generally, LMR, SLA, LAR, LARMR, leaf size and SSL decreased with irradiance, whereas leaf thickness and RMR increased with irradiance. Shade-grown plants possessed a larger crown area (Fig. 1c) than sun plants, but had a similar LAI (data not shown). Most plant variables showed a curvilinear response to light, with largest changes at the lowest light levels (3–12%), and a leveling-off at higher light levels (25–100%). However, the absolute responses to light were small for some variables (e.g. SMR, SSL and stem length).

Figure 1.

. Morphological responses to light for six rain-forest tree species. Species include two pioneer species (open symbols, Cecropia and Schizolobium), two intermediate species (grey symbols, Cedrela and Tachigali) and two shade-tolerant species (filled symbols, Cariniana and Theobroma). Figures refer to (a) allocation (root mass ratio, stem mass ratio, leaf mass ratio), (b) leaf display (leaf thickness, specific leaf area, leaf area ratio), (c) leaf size and crown area, and (d) stem elongation (specific stem length, stem length). Back-transformed logarithmic means are shown, except for plant variables strongly influenced by biomass (leaf size, crown area, SSL, stem length) for which back-transformed means at a common mass (5·0 g) are shown.


For most species RGR increased with increasing light availability at low irradiance, reached an optimum at 25–50% and declined again at full light (e.g. Tachigali, Fig. 2). Other species showed a relatively flat response curve (e.g. Bertholletia, Appendix). Surprisingly, the two pioneer species also showed a decline in growth at higher irradiance. For all species the LAR declined logarithmically with increasing irradiance. This trend was accompanied by an increase in NAR, which showed a logarithmic increase with irradiance for most species (e.g. Cariniana, Fig. 2), or had an optimum at intermediate light levels (25–50%) for others (e.g. Euterpe, Appendix). The RGR should be the product of NAR and LAR. In this study values of RGR and the product of LAR and NAR matched each other closely (Pearson = s, r = 0·96, P < 0·001, n = 89).

Figure 2.

. Responses of LAR (open circles, continuous line), NAR (open squares, broken line) and RGR (filled circles, continuous line) to light for Cecropia, a pioneer species, Tachigali, an intermediate species, and Cariniana and Theobroma, two shade-tolerant species. Means are given, as well as standard errors for LAR and RGR. Note that the scaling differs between graphs.


Between-species differences in growth were consistent among light environments. Species which realized the highest growth rate at low light also realized the highest growth rate in high light; the RGR of the species in 100% light was correlated with the RGR of the species in 3% light (Fig. 3a). This correlation was found for the other light environments as well (Fig. 3b).

Figure 3.

. (a) RGR of species at 100% light plotted against RGR at 3% (n = 14), (b) RGR of species at high light plotted against RGR of species at 3% light. Only regression lines are shown, for each light treatment one. Coefficients of determination and P-values are, respectively: 6% = 0·83***; 12% = 0·63***; 25% = 0·61***; 50% = 0·60***; 100% = 0·68***.

Fast-growing species have certain characteristics which enable them to outgrow the other species. At any given irradiance, both LAR and NAR were positively correlated with RGR (Fig. 4). At low irradiance interspecific variation in RGR was strongly correlated with the LAR. At higher irradiance levels this relation gradually disappeared. In contrast the relationship between NAR and RGR was weak at low irradiance levels and became gradually stronger at higher irradiance levels.

Figure 4.

. Relationship between LAR (filled circles), NAR (open circles) and RGR at (a) low (3%), (b) intermediate (25%) and (c) high (100%) light. Species means (n = 15), regression lines (LAR continuous line, NAR broken line), coefficients of determination and significance levels are shown.

For each variable a growth response coefficient can be calculated. This coefficient expresses the variation in RGR which is explained by variation in that variable, and can be calculated for each irradiance level. When the GRCs are plotted against irradiance (Fig. 5a) it can be seen that LAR had its highest effect on RGR at low irradiance, whereafter its influence logarithmically decreased (P < 0·05, r2 = 0·81, n = 6). In contrast, NAR had a small effect on RGR at low irradiance, whereafter it logarithmically increased (P < 0·01, r2 = 0·88, n = 6). The two lines cross at 10–15% irradiance, indicating that below this irradiance level LAR largely determines interspecific variation in RGR, and above this irradiance level NAR is the most important factor. The decreasing impact of LAR on RGR with increasing irradiance is the result of underlying patterns in SLA and LMR. At low irradiance SLA exerted a strong influence on RGR, but its influence declined dramatically at higher irradiance (Fig. 5b, P < 0·001, r2 = 0·99, n = 6). In contrast, LMR had little effect on RGR at low irradiance, but a larger impact at higher irradiance (P < 0·05, r2 = 0·83, n = 6).

Figure 5.

. Contribution of (a) LAR (filled circles) and NAR (open circles) and (b) LMR (open circles) and SLA (filled circles) to interspecific variation in RGR, as dependent on light. Growth response coefficients, standard errors and logarithmic regression lines (NAR, r2 = 0·88, P < 0·01; LAR, r2 = 0·81, P < 0·05; LMR, r2 = 0·83, P < 0·05; SLA r2 = 0·99, P < 0·001; n = 6 in all cases) are presented.


Regression of NAR against ln-transformed irradiance was significant for all but one species and yielded high coefficients of determination (mean r2 of the significant cases was 0·94, range 0·83–0·99). Using these regression lines a whole-plant light-compensation point could be calculated. Not much value should be given to the absolute value of this light-compensation point, as it is calculated on the basis of extrapolation. However, it might give an indication of the ranking of shade tolerance of the tree species. Theobroma, the understorey species, had the lowest light-compensation point, whereas Bellucia and Cecropia, the pioneer species, had the highest light-compensation points. The light-compensation point was positively correlated with maximum RGR (P < 0·05, n = 15, r2 = 0·36) (Fig. 6) and maximum NAR (P < 0·05, n = 15, r2 = 0·38) the species attained in one of the light treatments, but not with maximum LAR (P > 0·05, n = 15, r2 = 0·18). However, the significance of the relationships between RGR, NAR and light-compensation point hinges mainly on the two pioneer species Bellucia and Cecropia. If these species are excluded from the analysis, than the relationship disappears.

Figure 6.

. Relationship between maximum RGR attained in one of the light environments and whole-plant light-compensation point (LCP) of seedlings of 15 rain-forest tree species.



In this study most size- and architecture-related plant variables changed with size of the plant, whereas leaf- and water-balance related characteristics were more influenced by irradiance (Table 2). SSL appeared to be one of the variables most strongly influenced by ontogeny (Table 2). SSL declined with plant biomass as large-sized plants show more secondary thickening of the stem. Surprisingly, the LMR showed a modest increase with biomass. Yet, owing to a concomitant decrease in SLA, it resulted in a constant LAR over the biomass interval. From other studies it is known that LMR and LAR can decrease with plant size (e.g. Walters, Kruger & Reich 1993). In a literature compilation on growth of tropical tree seedlings it was found that all growth related plant traits were influenced by light and plant size (Veneklaas & Poorter 1998). Generally SMR increased with biomass, whereas LMR, SLA and LAR decreased. SLA and LAR were more strongly influenced by plant size than biomass allocation parameters. Sterck (1997) evaluated the relative importance of tree height and light environment for the architecture of large rain forest trees. He found that height was a far more important determinant of architecture than light.

Ontogenetic drift in plant variables is frequently observed for herbaceous species (e.g. Blackman & Wilson 1951; Evans 1972; Poorter & Pothmann 1992), but may be more pronounced in woody species (Körner 1994). As the life span of the seedling is longer than that of its modules, different turnover rates of plant components like roots, stems, twigs and leaves may lead to quite a different plant morphology over time.

Often it is difficult to disentangle whether a certain plant form is the result of a functional response of the plant to its environment or simply the consequence of plant size (Coleman, McConnaughy & Ackerly 1994; Huber & Stuefer 1997). Although ontogeny might be an important confounding factor for the evaluation of light responses of woody plants, there have been hardly any studies in which it is accounted for (but see Cornelissen 1993; Walters et al. 1993). One may compare plants over a standardized biomass growth interval (Van de Vijver et al. 1993) or at a single common biomass (Rice & Bazzaz 1989). In the latter case it is critical at which biomass species are compared, as species-specific ontogenetic trajectories of plant traits do not necessarily run parallel (Walters et al. 1993). Comparison at a common biomass offers an elegant solution for dealing with ontogenetic trends, but it has also a drawback in the sense that sometimes extrapolations need to be made to compare plants at a common biomass. Occasionally this might have led to an under- or overestimation of the parameter value.


In an aseasonal environment plant responses to light are governed by different resource constraints at each end of the light gradient. At low light plants enhance light interception by means of a high biomass allocation to leaves (Fig. 1a) and the formation of thin leaves with a high SLA, leading to a high LAR (Fig. 1b). Self-shading is potentially diminished by the formation of a wide crown (Fig. 1c). Nevertheless, this does not lead to a lower LAI. At high light plants reduce transpiration losses and increase carbon gain by making small-sized, thick leaves with a low SLA. Small-sized leaves have a thin boundary layer which allows for a better convective heat loss to the environment. In this way less transpiration is needed for cooling down the leaf in a high-light environment (Parkhurst & Loucks 1972; Givnish 1984). Leaf thickness is increased owing to the formation of several photosynthetically active parenchyma layers. In this way the photosynthetic capacity of the leaf is enhanced. Similar patterns have been observed for leaves of seedlings along a light gradient and for leaves of trees along a height gradient in the forest canopy (Oberbauer & Strain 1986; Poorter, Oberbauer & Clark 1995). Water uptake by the plant is accounted for by increased biomass investment in the roots, resulting in a high RMR (Fig. 1a) and a better balance between transpiring leaf surface and root biomass (LARMR).

Investment in stem mass was weakly dependent on light environment. Plants in the shade had a higher SSL (cf. Sasaki & Mori 1981), resulting in a higher stem length for shade plants compared to sun plants (Fig. 1d). Dependent on the length and steepness of the vertical light profile in the vegetation a plastic response in plant height can pay-off in terms of an increased light interception. It is often suggested that such a response is restricted to light-demanding species only, as they regenerate in a short-stature gap vegetation. By means of a plastic response in stem length, they may attain a position in the regrowing canopy. However, in this study also shade-tolerant species showed such a plastic response in SSL. In a Bolivian forest, mean canopy openness increases from 3·6% to 4·1% over a 1·2 m height interval (Arets 1998). An increased interception of light through a plastic response in height may lead to a little enhanced growth (Fig. 2, Kohyama 1991), or can put the seedling above its whole-plant light-compensation point.


For most species, RGR reached its optimum at intermediate light conditions (25–50%) above which it declined (Fig. 2). The RGR responses of plants to light can be explained by underlying patterns in LAR and NAR. A decline in LAR is offset by an increase in NAR, leading to a higher RGR at intermediate light intensities (Fig. 2). At full light, NAR does not compensate for a decline in LAR, leading to a lower RGR.

Generally it is expected that plant growth will increase with increasing irradiance. A possible explanation for the observed pattern might be that plants grown at full light suffered from water limitations or high soil temperatures. In this experiment, plants were watered every other day in addition to the natural rainfall. In addition, the full light treatment differed from the other treatments in the sense that it did not have a cover of shade cloth. Therefore the plants were less protected against wind and heavy rain. Nevertheless, the predictable, gradual patterns in LMR, SLA and LAR with increasing irradiance (Fig. 1) suggest that plant morphology at full light does not show an extreme response. Even if plants would have been irrigated continuously, a growth depression at high-light conditions would have been very likely for three reasons. First, very high irradiance levels may lead to irreversible damage to the photosynthetic system. For some species, like Tachigali, bleaching of the leaves was observed at full light. Second, high irradiance around midday leads to stomatal closure, and sometimes even to turgor loss and wilting of the leaves (Oberbauer 1985; Chiariello, Field & Mooney 1987). This may have such an impact, that even late in the afternoon, light-saturated photosynthetic rates can be considerably lower compared to the morning (Poorter & Oberbauer 1993). Some early successional species, like Schizolobium, have photonastic leaves to cope with excess radiation. Around noon the leaflets close themselves thus reducing intercepted radiation, only to open again in course of the afternoon. Third, high radiation loads require a larger biomass allocation to roots for water uptake to compensate for transpirational losses. Less biomass can therefore be invested in leaf material, which strongly reduces photosynthetic gain and potential growth rate (Körner 1991).

Many seedling growth studies consist of a few, unevenly spaced light treatments (e.g. Fetcher, Strain & Oberbauer 1983) so it is easy to miss the light level for maximum growth. Loach (1970) studied the growth of four temperate tree species at 3, 17, 44, and 100% light. For all species, maximal RGR was attained at 44% light, and at full light RGR was a little lower. In a study of the growth of Liriodendron under controlled light and water conditions, it was even found that maximum growth was realized at light levels as low as 12% (Holmgren 1996). Veenendaal et al. (1996) compared seedling growth of 15 West African tree species at various light levels. It was found that shade-tolerant species showed highest RGR at 16 or 27%, above which it declined. For the pioneer species optima were between 26 and 100%.


In tropical rain forest, light is one of the most limiting factors affecting plant growth and survival. It is often suggested that niche differentiation occurs and that rain-forest tree species separate along the light gradient (e.g. Denslow 1980). If this hypothesis holds, then species should be specialized for a certain range of the light gradient, at which they perform better than others. Implicitly, it is assumed that this is also reflected in the growth performance; in low light shade-tolerant species are hypothesized to outgrow pioneer species, whereas in high light the reverse is true. Nevertheless, in this study such a shift in ranking of growth performance was not observed; fast growers in the shade, are also fast growers in high light (Fig. 3b). This situation is comparable to growth patterns found for herbaceous species along a nitrogen-gradient. Fast-growing species from nutrient-rich habitats, outgrow the slow-growing species from nutrient-poor habitats, even under oligotrophic conditions (Fichtner & Schulze 1992; Keddy, Twolan-Strutt & Wisheu 1994).

A positive correlation between seedling growth in high and low light is also found in a number of other studies (e.g. Popma & Bongers 1988; Kitajima 1994; Osunkoya et al. 1994). Most of these studies were carried out at light levels comparable with, or a bit higher than the understorey (2–4%), but which were still well above the whole-plant light-compensation point of light demanding species. The positive relation between growth in low and high light may change if light levels are so low that they come close to the whole-plant light-compensation point of this species group. In a study on seedlings of six shade-tolerant tree species, there was no correlation between species growth above and below the whole-plant light-compensation point (Boot 1993). In Ghana tree species showed a reversal in seedling performance in high (65%) compared to low (2%) irradiance: whilst pioneer species realized highest growth rates in high light, they showed negative growth rates in low light whereas the shade-tolerant species maintained positive growth rates (Agyeman 1994 in Swaine et al. 1997).

The results of these studies are quite counter intuitive. With the exception of the last two studies, it appears that fast-growing, light-demanding species always grow better than shade-tolerant species. It might be that the shade tolerance of species is not related to growth but to persistence (Paccala et al. 1996). In this study the whole-plant light-compensation point is used as a measure of shade-tolerance. It indicates under what light levels plants still may persist. As such it is a measure which links light and potential survival. Species having a low LCP were indeed characterized by a low RGRMAX (Fig. 6, cf. Moad 1992 in Ackerly 1996) indicating that shade tolerance is related to low potential growth. Often a negative correlation is found between growth rate and survival. This applies for very small seedlings (Kitajima 1994), as well as for larger-sized saplings (Hubbell & Foster 1992; Kobe et al. 1995). It is suggested that species do not maximize their potential growth rate, but rather their realized biomass growth under field conditions (Kitajima 1996). Realized growth in the field is the result of two components: (1) biomass production through growth and (2) biomass losses as a result of herbivory, mechanical disturbance and shedding (Givnish 1988; Kitajima 1996). The relative importance of these two components differs between resource-rich and resource-poor habitats (Coley, Bryant & Chapin 1985). In a low-light environment, the second component is thought to be of overriding importance. Biomass loss is minimized by a low leaf turnover (King 1994) and by allocating resources to storage and defense. Risk of herbivory can be reduced by making thick, lignified leaves with a low SLA, but at the expense of a reduced potential growth. Indeed, it has been found that tree species which are less prone to herbivory are characterized by a high leaf toughness and low inherent growth rates (Cornelissen, Castro-Díez & Carnelli 1998).


Fast-growing species have certain characteristics which enable them to outgrow the other species. It depends on the light conditions used, which factors determine interspecific variation in RGR. Both LAR and NAR exert influence on RGR, but their importance changes along the light gradient. At low irradiance interspecific variation in RGR is strongly correlated with LAR, and at high irradiance with NAR (Fig. 5). A high LAR is of advantage in a low-light environment, where interception of light is of primary importance. A high NAR can only become of importance if there is sufficient light, so that plants can fully benefit from a high photosynthetic capacity. Shade-tolerant species are constrained by an inherently low photosynthetic capacity (Strauss-Debenedetti & Bazzaz 1996) and may suffer from photo inhibition (Chazdon et al. 1996) whereas pioneer species can realize high light-saturated photosynthetic rates at high irradiance. As there is a close relation between photosynthetic rates and NAR (Poorter & van der Werf 1998), it follows that at high-light species may differentiate in their NAR. Similar results were found by Veneklaas & Poorter (1998) in a literature compilation of seedling growth of 120 tropical tree species. Pioneer, intermediate and shade-tolerant species showed a consistent ranking of growth in low, medium and high light intensities, with the pioneer species always realizing the highest RGR and shade-tolerant species always realizing the lowest RGR. At low light, between-species differences in RGR were mainly owing to LAR and at high light they were mainly owing to NAR. For Australian rain-forest tree species (Osunkoya et al. 1994) the reverse pattern was found; at low light, interspecific variation in RGR was associated with NAR, whereas at intermediate light, it was associated with LAR. In another seedling study (Popma & Bongers 1988), variation in growth was associated only with NAR.

In a recent review of 55 studies on growth of herbaceous species, the majority from temperate and short vegetation (Poorter & van der Werf 1998), it was found that LAR explained on average 77% of interspecific variation in growth, whereas NAR only explained 24% of this variation. This is in contrast with the findings for woody seedlings (Poorter 1989; Veneklaas & Poorter 1998; present study) where NAR can play an important role, be it at higher light levels. Most comparative studies on plant growth are carried out at relatively low irradiance, which might explain why in some cases the importance of physiological traits is underestimated. Yet, it may also pinpoint to different selective forces in the habitats of herbaceous and woody species. In temperate grasslands, the herbaceous species tend to separate along a nutrient gradient whereas in tropical rain forests, tree species tend to separate along a light gradient. It is likely that these different environmental axes present different trade-offs and have led to different plant strategies.

Although NAR can be an important factor explaining interspecific variation in RGR of tree species, it only becomes of importance above 10–15% irradiance. Below this threshold level LAR largely determines interspecific variation in RGR (Fig. 6a). Typically, light availability on the rain-forest floor is much lower. In Costa Rica only 3% of the forest floor received irradiance of more than 15% (Clark et al. 1996) and in Guyana it was as little as 1% (Zagt 1997). Only in intermediate and large-sized gaps light availability can be higher (Chazdon & Fetcher 1984; Van der Meer & Bongers 1996). If one seeks to explain interspecific differences in growth in the forest understorey, one has to look for morphological components related to architecture, leaf area and leaf display. A large leaf area appears to be of prime importance. This is not so much attained by biomass allocation to leaves, as by the formation of thin leaves with a high SLA.


I am grateful to Miguel Cuadiay and Walter for taking care of the plants, and Yáskara Hayashida Oliver, Rene Aramayo and Pieter Zuidema for help with harvesting. Maarten Terlou of the Department of Image Processing and Design helped with the image analysis and Paul Westers gave statistical advice. La Universidad Técnica del Beni is acknowledged for use of the experimental garden and staff and personnel of PROMAB for their logistic support. Comments of Frans Bongers, René Boot, Peter Grubb, Marielos Peña, Hendrik Poorter, Mike Swaine, Marinus Werger and Roderick Zagt considerably improved the manuscript. Part of this work was financed by grant BO 009701 from the Netherlands Development Assistance.



This contains table 3