Are inter- and intraspecific variations of sapling crown traits consistent with a strategy promoting light capture in tropical moist forest?

Abstract

Background and Aims Morphological variation in light-foraging strategies potentially plays important roles in efficient light utilization and carbon assimilation in spatially and temporally heterogeneous environments such as tropical moist forest understorey. By considering a suite of morphological traits at various hierarchical scales, we examined the functional significance of crown shape diversity and plasticity in response to canopy openness.

Methods We conducted a field comparative study in French Guiana among tree saplings of 14 co-occurring species differing in light-niche optimum and breadth. Each leaf, axis or crown functional trait was characterized by a median value and a degree of plasticity expressed under contrasting light regimes.

Key Results We found divergent patterns between shade-tolerant and heliophilic species on the one hand and between shade and sun plants on the other. Across species, multiple regression analysis showed that relative crown depth was positively correlated with leaf lifespan and not correlated with crown vertical growth rate. Within species displaying a reduction in crown depth in the shade, we observed that crown depth was limited by reduced crown vertical growth rate and not by accelerated leaf or branch shedding. In addition, the study provides contrasting examples of morphological multilevel plastic responses, which allow the maintenance of efficient foliage and enable effective whole-plant light capture in shaded conditions under a moderate vertical light gradient.

Conclusions This result suggests that plastic adjustment of relative crown depth does not reflect a strategy maximizing light capture efficiency. Integrating and scaling-up leaf-level dynamics to shoot- and crown-level helps to interpret in functional and adaptive terms inter- and intraspecific patterns of crown traits and to better understand the mechanism of shade tolerance.

INTRODUCTION

The ecological importance of morphogenetic adaptation to light regime in determining plant acclimation and plant successional status is not as broadly recognized as physiological adaptation of photosynthesis. Yet, variations in light-foraging strategies and associated variations in tree morphology potentially play important roles in efficient light utilization and carbon assimilation in spatially and temporally heterogeneous environments (Vincent and Harja Asmara, 2008; Ishii and Asano, 2010). Selection for sapling performance in the strongly limiting light conditions of the forest understorey is likely to maximize light capture efficiency (Givnish, 1988; Pearcy and Valladares, 1999; Pearcy and Yang, 1998), defined as the ratio of mean light intercepted by leaves to light intercepted by a horizontal surface of equal area, averaged over the entire sky hemisphere (Delagrange et al., 2006). Saplings of shade-tolerant species are expected to have mono-layered crowns and shallower crown forms than saplings of pioneer species as a way to reduce self-shading and their whole-plant compensation point. This crown pattern was observed at the seedling or sapling stage across temperate tree species (Horn, 1971; Tucker et al., 1987; Valladares and Niinemets, 2008 – but see Lorimer, 1983) whereas studies conducted across tropical tree species (Poorter and Werger, 1999; Sterck et al., 2001 – but see Kohyama and Hotta, 1990) found the opposite pattern, namely a deeper crown among shade-tolerant species. Plant crowns perform multiple functions (including water transport, and mechanical support of leaves and reproductive organs) and therefore have multiple constraints on their form and function (Farnsworth and Niklas, 1995; Pearcy et al., 2004; Kennedy, 2010), challenging the interpretation and understanding of the role of crown trait variations in the relative performance of species along the light gradient. We contend that considering a suite of morphological traits at various hierarchical scales will enable us to gain insight into the functional meaning of crown trait variations and to improve our mechanistic understanding of differential species responses to shade.

Among species with an erect orthotropic main shoot and lateral branches, crown length dynamics is determined by the balance between growth of the crown tip and rise of the crown base as the lowest branches (or leaves in unbranched saplings) die. Leaf lifespan may drive crown length through its control on branch lifespan (Seiwa et al., 2006) (Fig. 1). Saplings of shade-tolerant tropical species are usually characterized by long leaf lifespan (LLS) and high leaf mass per area (LMA) as part of a conservative resource strategy (Westoby et al., 2002; Reich et al., 2003; Sterck et al., 2006, 2011). By contrast, pioneers and heliophilic species display rapid leaf turnover as a way to maintain leaf area in high-light environments and to increase light capture efficiency (Hikosaka, 2005). Hence, the deeper crown observed among shade-tolerant evergreen tropical saplings by Sterck et al. (2001) against Givnish’s assumptions might be explained by their longer LLS. The influence of LLS on crown depth and on survival in low light was investigated by Sterck et al. (2005) through a functional–structural plant growth model: the authors concluded that in a closed canopy, species with long LLS displayed a deeper crown and survived longer than species with short LLS which rapidly died because they were unable to replace dead leaves. This theoretical result has not yet, to our knowledge, been validated in a field comparative study. We conducted such a field study in French Guiana among saplings of 14 co-occurring tree species differing in light-niche optimum and breadth. We examined relationships between crown depth, LLS and crown vertical growth across species.

Fig. 1.

Determinants of crown depth at three organizational levels.

The second objective of the present study was to look at to what extent morphological plasticity of tropical tree saplings was consistent with a strategy promoting light capture efficiency and indirectly whole-plant carbon gain (Valladares and Niinemets, 2008). Givnish (1988) postulated that phenotypic plastic response may parallel interspecific adaptation to shade. We expected that such plastic adaptive adjustments would be more clearly expressed in species displaying a wide light-niche breadth given that phenotypic plasticity is a potential adaptation to environmental heterogeneity (Alpert and Simms, 2002). The difficulty in studying phenotypic plasticity is that potentially adaptive responses co-occur with passive responses that have neutral or negative implications on individual performances. Dudley (2004) argued that the first line of evidence for the adaptive value of plastic responses is knowledge of the biology of the organisms. Plant physiological ecology and botany provide a priori hypotheses about the causes of trait plasticity, and the functional or structural relationships among traits, and ultimately provide supporting evidence that an observed response positively impacts plant performance. We propose that a multivariate and multilevel approach towards phenotypic plasticity may significantly contribute to improve our understanding of species-specific plastic responses to shade.

As a first step, we applied the flow diagram described in Fig. 1 to analyse light-driven plasticity of crown depth and underlying traits within 14 tropical tree species. Givnish’s statements predict that crown depth would be shorter in shade than in high-light conditions as a result of the carbon and nutrient recycling of shaded leaves leading to the shortening of LLS and branch lifespan (BLS) and as a way of enhancing light capture efficiency. To our knowledge, reduced lifespan of shaded leaves and shoots has only previously been observed among temperate and deciduous light-demanding tree species (Seiwa et al., 2006), and in a perennial herb (Hirose et al., 1988). Two other processes are likely to drive in opposite directions intraspecific variations of crown depth along the light gradient. (1) An increase of LLS in shaded conditions presumably results in slower photosynthetic metabolism in shade, delaying leaf senescence. LLS was found to be negatively correlated with photosynthetic capacity within species in several studies ranging from herbaceous desert perennial to tropical tree species (Chabot and Hicks, 1982; Ackerly, 1999; Escudero and Mediavilla, 2003; Vincent, 2006). This increase of LLS in shade may extend BLS and indirectly crown length (for a constant number of branches). (2) There is a lower vertical growth rate of the crown in shade following the reduction in leaf emission rate of the main stem and of branches, negatively impacting crown depth.

As a second step, we extended this multilevel approach to other crown traits driving light capture while focusing on a subset of four species. These species display a similar monopodial growth model and so a common set of underlying traits but a contrasted degree of shade tolerance and light-niche breadth. We examined plastic responses of interrelated traits (Supplementary Data, Fig. S1) underlying crown slenderness, leaf dispersion and total leaf area.

Plasticity in crown slenderness (ratio between crown depth and crown width).

Variations in crown slenderness in shade can result from two processes: (1) a reduction in relative crown depth analysed above and (2) adaptive adjustments in relative growth of horizontal branches and of the vertical main stem. We expected that low-light saplings of less specialized species would invest proportionally more in horizontal crown growth than in vertical crown growth as a way to favour light capture (Sterck, 1999). The resulting ‘plate-shaped’ crown is assumed to increase light capture efficiency in the light-limited understorey (Kohyama and Hotta, 1990; Valladares et al., 2002; Sterck et al., 2003).

Plasticity in leaf dispersion.

Spatial aggregation of leaves has been shown to be a strong driver of light interception efficiency throughout a large database of digitized plants (Duursma et al., 2012). Leaf dispersion was estimated by branch spacing and internode length. Shorter branch spacing is likely to impact negatively on light interception by increasing self-shading but to enable a reduction in leaf display costs. King et al. (1997) observed a reduction in first-order branch spacing resulting from a reduction in internode length within understorey saplings of 14 tropical tree species in Panama under more shaded conditions.

Plasticity in total leaf area.

Variations in total leaf area were analysed in relation to the variations in the length of leaf cohorts, in the number of branches and in the unit leaf area. We expected a significant reduction in the length of leaf cohorts and in the number of branches, and consequently a reduction in total leaf area, in shaded conditions within light-demanding species as a result of the reduction in leaf emission rate coupled with their shorter LLS. The ability of a given species to survive in shaded conditions may be directly determined by its ability to maintain a sufficient leaf area (Lusk, 2002). King (1994) argued that saplings of shade-intolerant species are excluded from typical understorey sites because their short LLS necessitates higher emission rates to maintain their canopies. Consequently, we interpreted plasticity of total leaf area as a non-adaptive response to low-light conditions.

In summary, we examined the relationships between crown depth, LLS and crown vertical growth across 14 tropical tree species and three light regimes. We expected that: (1) shade-tolerant species would display a deeper crown than heliophilic species as a consequence of their longer LLS and (2) that shaded trees would display a shorter crown than high-light trees because of shortened LLS. In addition, for a subset of four species differing in light-niche position and breadth, we examined the plasticity of a suite of crown structure and leaf traits and discussed the potential impact of these adjustments on light capture and on relative performances of species along the horizontal light gradient of tropical forests.

MATERIALS AND METHODS

Study site and field measurements

The study was conducted in a lowland tropical rain forest at the Paracou experimental site (5°18′N, 52°55′W) in French Guiana. Rainfall averaged 2875 mm year⁻¹ over the period 1986–2005 with a 3-month dry season (<100 mm per month) from mid-August to mid-November. The forest shows the high species diversity typical of tropical rain forest and a very high proportion of rare species (Ter Steege et al., 2006; Rutishauser et al., 2011). The dominant families at the site include Fabaceae, Chrysobalanaceae, Lecythidaceae, Sapotaceae and Burseraceae. The 14 non-pioneer co-occurring species studied are common forest species in French Guiana and account for 27 % of the total tree population (>10 cm diameter at breast height) at the Paracou experimental site.

To evaluate species-specific responses to different light regimes, an extensive search throughout the Paracou experimental site was conducted to identify suitable saplings (0·5–3 m tall) in all light regime classes. These saplings were selected outside seasonally flooded areas and any obviously resprouted stem was excluded. Overall, 41–76 saplings per species (total 844) were selected, tagged and mapped. All saplings and their light environments were measured annually from 2007 to 2009 (or from the date of first encounter, after 2007).

Light measurement

The light environment of each sapling was evaluated during each census by two observers using a light regime visual estimate based on the structure of the vegetation above and around the sapling. We used a scoring system similar to that of Clark and Clark (1992) adapted to suit the forest structure at Paracou where 1 = no direct light, dense understorey; 2 = light understorey (some lateral light due to nearby gap, or thin upper canopy layer); 3 = significant direct illumination associated with position either on the edge of a large gap or well inside a small gap; 4 = abundant vertical illumination (large gap centre, track side). The mean of the two observers’ scores was recorded for each census and the average light environment for each sapling was described by calculating the mean light index (LI) value for all the different censuses. The reliability of this index was assessed for a subset of individual plants by comparing values with two other methods (hemispherical photography and quantum photosynthetically active radiation sensor) detailed in Laurans et al. (2012) and found to be acceptable. Although these other methods were potentially more accurate, they were unsuitable for use with large data sets over rugged terrain.

Species description

Tropical trees have been classified into 23 different growth models by Hallé et al. (1978) based on differences in axis orientation, growth rhythmicity and terminal or lateral position of flowers. The 14 species exhibited contrasting architectures and developmental patterns (Table 1 and Fig. 2) and can be divided into four categories. Three species (Symphonia sp1, Virola michelii and Qualea rosea) with a monopodial main stem differentiate plagiotropic and monopodial lateral branches from the orthotropic leader stem. These species are assigned to Massart’s model in Hallé’s classification. The development of Oxandra asbeckii conforms to Roux’s model with a monopodial orthotropic stem growing continuously as with its plagiotropic branches whereas the development of Tachigali melinonii corresponds to Rauh’s model with differentiation of the monopodial stem growing rhythmically. Six species (Bocoa prouacensis, Dicorynia guianensis, Eperua falcata, Eperua grandiflora, Licania alba, Lecythis persistens) have a main stem formed sympodially from plagiotropic shoots and conform to Troll’s model. Among them, the saplings of two species (E. falcata, B. prouacensis) display axes that are more or less plagiotropic. For these species, leader stem and branches cannot be distinguished from one another within the developing crown. Thus, the highest shoot was defined as the top of the main stem and other axes branching from this main stem were treated as lateral branches. At a later stage, a stem becomes visible as the axis that is most upright and that is thicker than the other axes which are eventually shed or develop more horizontally. In conformity with Aubreville’s model (Fisher and Hibbs, 1982), saplings of three species (Gustavia hexapetala, Pradosia cochlearia and Sextonia rubra) develop sympodial, plagiotropic branches. Pradosia cochlearia was the only species which modified its entire physiognomy along the light gradient. In the shaded understorey, it exhibited a plagiotropic and polyarchic development whereas in gap conditions, it exhibited an orthotropic and hierarchical development.

Table 1.

List of study species with abbreviations used (initial of genus name followed by initial of species name) in the figures, diameter range (min.–max.) and architectural model as defined by Hallé et al. (1978), leaf type (C = compound, S = simple)

Family	Species	Abbreviation	Diameter range (mm)	Architectural model	Leaf type	Phyllotaxy	POP-RESP	Degree of specialization	Hmax
Fabaceae	Bocoa prouacensis	BP	5·6–20	Troll	S	Alternate	–0·03	1·9	34
Fabaceae	Dicorynia guianensis	DG	6–20·2	Troll	C	Alternate	0·00	1·2	52
Fabaceae	Eperua falcata	EF	5–18·5	Troll	C	Alternate	0·03	1·7	44
Fabaceae	Eperua grandiflora	EG	5·2–17·2	Troll	C	Alternate	–0·06	2·1	42
Lecythidaceae	Gustavia hexapetala	GH	8·3–18·3	Aubréville	S	Alternate	0·04	1·8	20
Chrysobalanaceae	Licania alba	LA	5–17·2	Troll	S	Alternate	0·00	1·2	31
Lecythidaceae	Lecythis persistens	LP	8·7–19·3	Troll	S	Alternate	–0·02	1·3	37
Annonaceae	Oxandra asbeckii	OA	5·9–17	Roux, Troll ?	S	Alternate	–0·08	2·2	18
Sapotaceae	Pradosia cochlearia	PC	5·6–17·2	Aubréville	S	Opposite	–0·05	1·6	49
Vochysiaceae	Qualea rosea	QR	5·5–17·9	Massart?	S	Opposite	0·03	1·6	46
Lauraceae	Sextonia rubra	SR	4·8–20·2	Aubréville	S	Alternate	0·00	1·4	44
Clusiaceae	Symphonia sp1	SS	6–16·6	Massart	S	Opposite	0·00	2·6	26
Fabaceae	Tachigali melinonii	TM	5·4–16	Rauh/Petit	C	Alternate	0·08	3·5	35
Myristicaceae	Virola michelii	VM	3·1–23·2	Massart	S	Opposite	0·07	2·8	41

Light-related niche traits are indicated: POP-RESP is a measure of the correlation between abundance and degree of canopy openness, which reflects the niche optimum; degree of specialization reflects the species sensitivity to canopy openness and indicates species niche breadth, and H_max (m) indicates adult stature.

Fig. 2.

Architectural pattern of the 14 studied species. Each green foliage symbol represents a leaflet of compound leaves or a pack of simple leaves. Photographs show the foliage of the species named in bold.

The sampled species can be divided into two groups of leaf type (compound leaves vs. simple leaves; Table 1). We defined as ‘leader stem’ the axis responsible for height growth and as ‘branch’ any other axis within the crown.

Light-niche characterization

Quantitative measures of light-niche parameters were estimated independently from the dataset and results of Vincent et al. (2011) and Laurans et al. (2012). Estimation of light-niche optimum was based on the correlation between species abundance and degree of canopy openness. This index is termed POP-RESP (population response to canopy disturbance) and ranged from −0·08 to 0·08 for the subset of species considered in the present study. Of the 14 species studied here, five had a value greater than 0, indicating that species abundance decreased with distance from disturbed areas, five had values below 0, indicating an opposite trend, and four had values not significantly different from 0, suggesting that disturbance had no – monotonic – effect on sapling abundance. The second light-niche parameter reflects the species sensitivity to canopy openness and indicates light-niche breadth. It is termed degree of specialization in the current study. Lower degrees of specialization values were indicative of a broader light-niche.

We excluded pioneer and truly shade-tolerant species and selected 14 species with intermediate light requirement. They exhibited a variety of light-niche breadths, with degree of specialization ranging from 1·2 to 3·5 (Table 1 and Supplementary Data Fig. S2). We analysed in more depth the morphological plasticity of four species: Qualea rosea, Virola michelii, Oxandra asbeckii and Symphonia sp1.

We evaluated a series of plant traits at three different levels of organization: whole-plant crown, axis and leaf (listed in Table 2). We report the median value of plant traits per LI class (the mean LI was rounded to the nearest integer) and per species and the plasticity across light environments of those traits.

Table 2.

List of traits used to describe the light niche of species and the morphology and the light environment of saplings

Level	Trait (units)	Abbreviation
Leaf	Unit leaf area (cm2)	ULA
	Leaf mass per area (g m−2)	LMA
	Leaf lifespan (d/months)	LLS
	Leaf loss rate (year−1)	LLR
	Leaf emission rate (year−1)	LER
Axis	Branch lifepan (months)	BLS
	Internode length (cm)	IL
	Leaf cohort length (leaves)	LCL
	Stem slenderness	HD
Crown	Branch spacing (cm)	BS
	Number of branches	BN
	Depth (cm)	CD
	Relative depth (%)	RCD
	Total leaf area (cm2)	TLA
	Slenderness	CS
	Lateral growth (cm year−1)	CLG
	Vertical growth (cm year−1)	CVG
	Crown rise (cm year−1)	CR
Plant	Diameter growth rate (mm year−1)	GRDIA
	Light index class	LI
Species	Light-niche optimum	POP-RESP
	Light-niche breadth	DEGREE OF SPECIALIZATION

Whole-plant functional traits

Diameter growth rate (GRDIA).

Stem diameter was measured at a marked position on the stem ∼20 cm from the ground and in two orthogonal directions. Vernier calipers were precise to within 0·1 mm. GRDIA was assumed to be linear over the study period and was calculated as:

$$\text{GRDIA\hspace{0.5em}}=\left[\left(G_{\text{2}}-G_{\text{1}}\right)/\left(t_{\text{2}}-t_{\text{1}}\right)\right]\times \text{365}$$

where G₁ and G₂ are diameter (mm) at t₁ (date of first census) and t₂ (date of last census).

Crown and axis traits

The number of leaves and branches (branch number, BN) of the whole plant was counted annually. The height of the lowest leaf or branch, the maximum crown width and the crown width perpendicular to it were also measured annually.

Total leaf area (TLA, cm²).

This was estimated as the product of the unit leaf area (ULA, cm²) and the total number of leaves (BN).

Crown depth (CD, cm).

Crown depth was defined as the difference between plant height and the height of the lowest leaf or branch. More than half of the saplings of D. guianensis, S. rubra and E. grandiflora were unbranched, so for these species crown represents most often leaf number and arrangement whereas for the other species crown represents branch number and arrangement.

Relative crown depth (RCD).

RCD was calculated as the ratio between crown depth and plant height.

Crown slenderness (CS).

CS was defined as the ratio between crown depth and maximum crown width.

Mean branch spacing (BS, cm).

This was calculated as the ratio between crown length and the number of tiers for V. michelii and the number of branches for the other species. It is considered a proxy of vertical self-shading within the crown.

Extension ratio: differential growth between main stem and first-order branches.

The length of the main stem and of an upper plagiotropic branch was measured at the beginning and the end of the experiment. We then calculated and compared the growth rate of each axis over the sampling period [crown vertical growth (CVG) for the main axis and crown lateral growth (CLG) for the upper plagiotropic branch in cm year⁻¹] to evaluate the strength of apical control in contrasted light environments. Apical control regulates the amount of elongation and of diameter growth of branches; there is a wide range of levels of apical control under different conditions both between and within individual plants (Wilson, 2000). A preferential investment in horizontal growth is likely to reduce self-shading.

Crown rise (CR, cm year⁻¹).

From the repeated measures of the height of the lowest (h) branch or leaf, we computed a rate of branch shedding (or leaf shedding where the main stem was unbranched) over the sampling period (t₂ − t₁) as follows:

$$\text{CR\hspace{0.5em}}=\left[\left(h_{\text{2}}-h_{\text{1}}\right)/\left(t_{\text{2}}-t_{\text{1}}\right)\right]\times \text{\hspace{0.5em}365.}$$

Branch lifespan (BLS, months).

LLS need not correlate with branch lifespan and hence with crown depth especially if branches display indeterminate growth.

As we did not measure the rate of branch loss or branch production, we estimated it from crown rise (CR), branch number (BN) and branch spacing (BS). The model applied is a biological application of a widely used law in queuing theory called Little’s law (Little, 1961), which states that the time-averaged number of arrivals in a queuing system, 1, is equal to arrival rate λ times the average sojourn time w:

$$\text{BLS\hspace{0.5em}}=\text{\hspace{0.5em}BN\hspace{0.5em}}/\left[\text{CR\hspace{0.5em}}\times \left(\text{1}/\text{BS}\right)\right].$$

We computed the Pearson correlation coefficient between BLS and LLS. We excluded from this analysis three species (Bocoa prouacensis, Eperua falcata, Eperua grandiflora) that displayed a bending main stem and for which, consequently, branch shedding is not necessarily correlated with crown rise.

Stem slenderness (HD).

This was defined as the ratio between stem height and stem diameter.

Mean internode length (IL, cm).

IL was estimated for the shoot of the leader stem developed during the census period. IL was calculated as the ratio between the length and the number of nodes of this shoot portion.

Leaf cohort length (LCL).

Leaf censuses were conducted on the main axis and on three first-order branches of each sapling. A leaf cohort was defined from the youngest leaf (fully expanded at the first census) to the oldest leaf found at the base of the axis. In the first census (conducted either in June 2007, November 2007 or February 2008), a record was made of the number of leaves on each monitored axis and the position of the youngest leaf in the cohort was marked using coloured adhesive tape. The number of leaves remaining in each sequence was further recorded (in July 2008 then in July or November 2009), yielding a sampling period of 5–30 months depending on axis lifetime. LOW_LCL, MED_LCL and HIGH_LCL refer to leaf cohort length of, respectively, lower, middle and upper first-order branches.

Leaf traits

Leaf lifespan (LLS, months), leaf loss rate (LLR, leaves year⁻¹), leaf emission rate (LER, leaves year⁻¹).

LLR and LER were estimated from measures of leaf cohort length for the main axis, respectively as the ratio of the number of dead and new leaves within the sampling period. We estimated LLS by applying Little’s law (Little, 1961) as in numerous previous studies (Southwood et al., 1986; Ackerly, 1996; Wright et al., 2002; Navas et al., 2003). This model assumes a steady-state system, meaning that the axis must be in a process of active leaf emission and loss. Because of the discrete leafing (flushes) of some species (E. falcata, E. grandiflora), LLR was used instead of LER. LLS was estimated per individual plant as the ratio of leaf number (leaf cohort length, LCL) to LLR on the main axis of each sapling:

LLS = LCL/LLR

Final size of the leaf population used in the analysis was 8625 leaves, with an average of 616 leaves per species.

LMA.

We collected five punches in fully expanded leaves located in the periphery of the upper crown half, between the main veins of leaves with a core of standardized area (diameter = 16 mm) in July 2008. LMA (g m⁻²) was calculated from leaf punch dry mass (oven-dried for 96 h at 65 °C) and punch area.

Unit leaf area (ULA, cm²).

A sample of leaves (1–4) was collected in a sub-sample of saplings per species. After leaf scanning, leaf area was quantified with the software Image J (http://rsb.info.nih.gov/ij/index.html). Unit leaf area of a given sapling was calculated as the mean of leaf area.

Sapling size effect

Sapling size might be an important confounding factor for the analysis of plant responses to light availability. Consequently, in spite of the small range of sapling size investigated, we tested the occurrence of a size effect on the whole set of measured traits. For most species, we observed a significant effect of stem diameter on crown dimensions, vertical growth, stem diameter increment, unit leaf area, number of leaves and number of branches. The significance of the size effect on the other traits depended on species. The size effect was controlled for by fitting a species-specific linear relationship between trait values, LI and stem diameter. If the size effect was significant (P < 0·05; 41 % of cases), we corrected for it in the following way. We predicted the expected trait value at species median diameter value and observed LI. Individual trait values were then recomputed by adding residuals of the original dataset to the model predictions. For all analyses we used these size-standardized values instead of the raw observed trait values.

Plasticity in functional traits

Median values of functional traits were computed for three light environments (low-light LI = 1, medium-light LI = 2, high-light LI = 3) after rounding each individual LI score. As the 14 species were not represented in LI = 4, this light class was excluded from the analysis. Plasticity in functional traits was quantified by the plasticity index (PI):

$$\text{PI\hspace{0.5em}}=max-min$$

where max represents the maximum median value of a functional trait across light classes, and min represents the minimum median value of a functional trait across light classes.

The significance of the plasticity index was tested by a Kruskall–Wallis test on functional trait values observed in the three light classes. Standard deviations were computed for the plasticity index estimate using bootstrap resampling [Boot package in R software (R Development Core Team, 2011)].

Analysis of interspecific and intraspecific variation

We carried out multiple regression analysis to estimate the respective influence of LLS, LER of the lowest branch and main axis vertical growth (CVG) on inter- and intraspecific variations of relative crown depth (RCD ∼ LLS + CVG + LER).

The general pattern of inter- and intraspecific variation of crown shape was evaluated by principal components analysis (PCA) of species median values of crown traits measured in low-, medium- and high-light conditions. POP-RESP and LI class were later correlated with the PCA axes.

We restricted more in-depth analysis of morphological plasticity to a subset of four species (Q. rosea, V. michelii, S. sp1, O. asbeckii). We ran PCA successively for these four species on data at the individual level. LI was later correlated with the PCA axes.

Statistical analyses were performed with the R software (R Development Core Team, 2011) on untransformed values of traits (except for the size effect correction as explained in the previous paragraph).

RESULTS

General trend of inter- and intraspecific variation of crown shape

The PCA performed with species- and LI-specific median values for the leaf, crown and whole-plant showed orthogonality between species and LI effects (Fig. 3). The first PCA axis explained 40·5 % of trait variations and reflected the light gradient whereas the second axis explained 27·7 % of trait variation and reflected the shade-tolerance gradient with positive values corresponding to the most shade-intolerant species. Countergradient variation of traits produced the following patterns: the most shade-tolerant species displayed higher LMA, longer LLS, deeper crown and lower diameter growth rate. Shade saplings displayed longer LLS, but lower branch number, lower LMA, a shallower and less slender crown than sun trees and lower diameter growth rate.

Fig. 3.

Plot of PCA ordination diagram showing traits (arrows in A). B and C represent projection of points (median value by species and by LI class) in the trait space: B, points are grouped by LI class; C, points are grouped by species class. Ellipses summarize the scatterplot. Abbreviations of traits are given in Table 2.

Determinants of inter- and intra-specific variations of relative crown depth

PCA (Fig. 3) and Pearson coefficient analysis (Pearson’s r = −0·54, P = 0·043) showed that more shade-tolerant species tended to have deeper crowns. Across species, multiple regression analysis showed that relative crown depth in low light was driven by LLS (F = 6·82, P = 0·028) and not by either crown vertical growth (F = 0·07, P = 0·79) or leaf emission rate of the lowest branch (F = 0·82, P = 0·386). Relative crown depth in high-light conditions was not predicted by any of the three traits tested. We verified that LLS correlated positively with BLS among the set of 11 species that exhibited a clear hierarchical branching pattern (Pearson’s r = 0·62, P = 0·041).

Within species, multiple regression analysis showed the significant positive effect of crown vertical growth on RCD for the four species displaying a reduction in crown depth (D. guianensis, Q. rosea, T. melinonii, V. michelii) in the shade and no significant correlation between LLS and RCD except for Q. rosea (Table 3).

Table 3.

Within-species multiple regression predicting relative crown depth from leaf lifespan, crown vertical growth and leaf emission rate of the lowest branch

Species	LLS			Crown vertical growth			Leaf emission rate
	Coef	F	P	Coef	F	P	Coef	F	P
BP	3·00E–04	0·52		0	0·13		0	0·4
DG	–0·0012	0·03		0·007	8·59	*	0·007	5·79	*
EF	0·0011	0·08		0·003	1·63		0·003	0·29
EG	–3·00E–04	0·4		0	0		0	0·01
GH	5·00E–04	0·2		–0·002	0·13		–0·002	0·01
LA	–0·0021	3·73	#	0·003	1·78		0·003	1·67
LP	2·00E–04	0·73		0	0		0	0
OA	–2·00E–04	0·26		0·001	0·21		0·001	0·06
PC	8·00E–04	0·26		0·005	0·83		0·005	0·06
QR	–0·0013	10·41	**	0·005	18·37	***	0·005	0
SR	0·0029	0·01		0·008	5·35	*	0·008	0
SS	7·00E–04	1·53		0·002	6·24	*	0·002	3·94	#
TM	–0·0015	1·29		0·004	6·82	*	0·004	1·7
VM	–9·00E–04	1·11		0·004	10·43	**	0·004	0·04

Significance levels are shown with: #P < 0·1, *P < 0·05, **P < 0·01, ***P < 0·001. Underlined species names indicate a significant effect of light index on relative crown depth.

Patterns of morphological plasticity in response to canopy openness within four monopodial species

Crown slenderness (ratio of crown depth to crown width) of Q. rosea and V. michelii changed significantly across light conditions. RCD declined in shaded conditions from 0·7 to 0·4 for Q. rosea and from 0·6 to 0·2 for V. michelii. Branch spacing and internode length were significantly reduced in low-light conditions (Table 4) for these two species (Fig. 4). These species displayed contrasting variation in other traits. Extension rates of Q. rosea were the same between main stem and upper branches in high-light conditions while extension rate on upper branches was higher than extension rate on main stem in low-light conditions (Table 5). Total leaf area of Q. rosea was constant along the light gradient. Length of leaf cohorts on upper and middle branches was higher in low-light conditions than in high-light conditions. LLS of Q. rosea was significantly higher in low-light conditions. Extension rates of V. michelii were the same between main stem and upper branches in low-light whereas extension rate was higher on main stem than on upper branches in high-light conditions (Table 5). Total leaf area was significantly reduced in low-light conditions (Table 4). The number of branches and tiers for a given plant size and the length of leaf cohorts (number of leaves emitted per terminal meristem per flush) were significantly reduced in low-light (Fig. 4). LLS was not correlated with the light gradient. Unit leaf area of V. michelii decreased in high-light conditions.

Table 4.

Plasticity index of crown, axis and leaf traits

Trait	Variable	Species
		O. asbeckii	Q. rosea	S. sp1	V. michelii
BS	Plasticity index	0·000	2·682	0·000	6·171
	s.d.	NA	0·877	NA	1·441
	P		**	#	**
	Response sign	NA	+	NA	+
CS	Plasticity index	0·000	0·440	0·000	1·000
	s.d.	NA	0·088	NA	0·166
	P		***	#	***
	Response sign	NA	+	NA	+
HD	Plasticity index	0·000	0·000	0·000	0·000
	s.d.	NA	NA	NA	NA
	P			#
	Response sign	NA	NA	NA	NA
IL	Plasticity index	0·000	2·950	2·880	6·520
	s.d.	NA	0·802	0·770	1·591
	P	#	***	***	**
	Response sign	NA	+	+	+
LMA	Plasticity index	11·075	14·510	9·995	15·185
	s.d.	2·759	2·099	3·851	4·479
	P	***	***	***	***
	Response sign	+	+	+	+
LLS	Plasticity index	34·660	46·770	46·177	0·000
	s.d.	NA	13·333	20·256	NA
	P	*	**	***
	Response sign	–	–	–	NA
BN	Plasticity index	3·029	8·000	0·000	4·393
	s.d.	0·849	2·289	NA	1·198
	P	*	***		**
	Response sign	+	+	NA	+
HIGH_LCL	Plasticity index	0·000	8·000	0·000	0·000
	s.d.	NA	2·575	NA	NA
	P		**	#
	Response sign	NA	–	NA	NA
LOW_LCL	Plasticity index	0·000	0·000	0·000	4·500
	s.d.	NA	NA	NA	1·693
	P				*
	Response sign	NA	NA	NA	+
MED_LCL	Plasticity index	3·500	0·000	0·000	3·000
	s.d.	1·333	NA	NA	1·776
	P	*			*
	Response sign	+	NA	NA	+
RCD	Plasticity index	0·000	0·312	0·000	0·378
	s.d.	NA	0·077	NA	0·061
	P		***		***
	Response sign	NA	+	NA	+
TLA	Plasticity index	0·000	0·000	0·000	1547·122
	s.d.	NA	NA	NA	556·807
	P			#	*
	Response sign	NA	NA	NA	+
ULA	Plasticity index	4·229	0·000	3·051	1·242
	s.d.	0·959	NA	0·919	0·289
	P	***		**	**
	Response sign	–	NA	+	–
CVG	Plasticity index	0·000	23·288	22·308	36·305
	s.d.	NA	9·135	4·070	7·525
	P		***	***	***
	Response sign	NA	+	+	+

Monotonic increase or decrease of median values with light level are indicated by ‘+’ and ‘–’, respectively. Standard deviation (s.d., calculated by bootstrapping) of Plasticity index is given. Significance levels of this test are shown with: #P < 0·1, *P < 0·05, **P < 0·01, ***P < 0·001. Abbreviations of traits are given in Table 2. NA, not applicable.

Fig. 4.

Plot of principal-component analysis based on individual trait values for four species. Abbreviations of species name and traits are given in Tables 1 and 2, respectively. Ellipses summarize the scatterplot.

Table 5.

Significance of ANOVA comparing growth of the main stem (CVG) versus growth of an upper branch (CLG) in low-light (LI = 1) and high-light (LI = 3) conditions

Esp	Ratio vert_growth/lat_growth	LI1		Ratio vert_growth/lat_growth	LI3
		Response	P		Response	P
BP	1·21	vert = lateral		0·21	vert = lateral
EF	1·01	vert = lateral		2·03	vert>lateral	#
EG	2·17	vert>lateral	*	1·66	vert = lateral
GH	0·79	vert = lateral		2·04	vert = lateral
LA	3·26	vert>lateral	*	2·20	vert>lateral	*
LP	3·60	vert>lateral	*	1·28	vert = lateral
OA	1·51	vert = lateral		2·46	vert>lateral	*
PC	1·33	vert = lateral		0·99	vert = lateral
QR	0·55	vert<lateral	**	1·28	vert = lateral
SR	1·09	vert = lateral		7·32	vert = lateral
SS	1·23	vert = lateral		2·74	vert>lateral	***
TM	1·41	vert = lateral		2·26	vert>lateral	*
VM	1·28	vert = lateral		6·21	vert>lateral	**

Significance levels are shown with: #P < 0·1, *P < 0·05, **P < 0·01, ***P < 0·001.

Symphonia sp1. displayed a variation in crown slenderness across light conditions but no variation in crown depth (Fig. 4). This change in crown form could result from the change in extension ratio: like V. michelii, extension rates were the same in upper branches and main stem in low-light conditions whereas the latter dominated in high-light conditions (Table 5). Unit leaf area, total leaf area, branch spacing and internode length decreased significantly in low-light conditions (Table 4).

Oxandra asbeckii displayed no significant variation in crown slenderness or crown depth across light conditions (Fig. 4). The extension rate of the main stem was higher than that of branches in high-light conditions but equivalent in low-light conditions. Branch spacing and total leaf area were the same in low- and high-light conditions. LLS and unit leaf area increased in low-light conditions.

DISCUSSION

Does leaf longevity explain among- and within-species patterns of crown depth?

As expected, we did not find parallel patterns between shade-tolerant and heliophilic species and between shade and sun plants. Plastic and evolutionary responses to shade were not congruent. Shade-tolerant species did not exhibit a shallow and flat crown but shade saplings did so. This counter-gradient variation of crown traits is consistent with previous studies conducted in tropical forests (Poorter and Werger, 1999; Sterck et al., 2001).

As hypothesized we found that across species, variations of crown depth result from the variation in LLS. The significant correlation between LLS and BLS explains why in spite of a faster crown vertical growth, the number of branches and the relative crown depth of heliophilic species is lower than those of the most shade-tolerant species. Hence high-light species tended to have shorter crowns. Correlation between LLS and BLS has been previously reported in deciduous temperate trees (Seiwa et al., 2006; Shirakawa and Kikuzawa, 2009). Seiwa et al. (2006) also provided evidence of a close relationship between BLS and successional status in the families Betulaceae and Fagaceae: BLS was shorter in early-successional species than in late-successional species. Such coordination between LLS and BLS provides support to the ‘fast–slow’ plant economics spectrum suggested by Reich (2014).

The deep crown of shade-tolerant species proceeds from a high longevity of resource-acquiring tissues, which compensates for limited carbon fixation rate. At the whole-plant level, long LLS enables accumulation of a large foliage area, which directly enhances carbon gain under low light (King, 1994; Lusk, 2002). In the study by Lusk (2002) which dealt with saplings in temperate rain forest, this feature did not result from high allocation to leaf but rather from the very low leaf loss rates. Furthermore, reduction in tissue loss reduces carbon demand for growth (Walters and Reich, 1999) and allows a greater allocation to other processes that directly contribute to stress resistance (defence compounds, carbohydrate storage) (Fine et al., 2006; Poorter and Kitajima, 2007). In contrast, a much shorter leaf lifetime is consistent with a strategy optimizing resource acquisition per unit leaf area because (1) rapid tissue turnover maintains the leaf area in favourable light conditions, where resource gain per unit leaf area can be maximized and (2) it increases the efficiency of resource deployment, as a result of nutrient recycling (Seiwa et al., 2006). However, as most of the carbon and more than half of the nutrients in individual leaves and shoots are lost at senescence (Seiwa et al., 2006), this strategy is only beneficial in high light conditions or in the case of a steep vertical light gradient between the base and the top of crown and probably in relatively fertile conditions (Richardson et al., 2010).

Thus, variation of crown depth across species showed the opposite trends to those predicted by Givnish (1988) stating that shade-tolerant species would show adaptation to low light via flatter (and more sparsely foliated) crowns, which would reduce self-shading. Conversely, our results suggest that leaf traits of shade-tolerant species may enhance whole-plant energy capture, whole-plant carbon gain and indirectly survival in low-light conditions through the accumulation of high leaf area. Clearly crown depth is a crude surrogate of self-shading and a deep crown is not necessarily incompatible with a low level of self-shading especially in the understorey where light may less preferentially come from above and in the humid tropics where much of the incoming light year round is diffuse because of the frequently cloudy conditions. By considering the other architectural determinants of self-shading, we might show that the two selective pressures outlined above (prolongation of LLS and reduction in self-shading) are not as strongly opposed as imagined. Trees can avoid self-shading by other morphological traits such as phyllotaxy, leaf shape, and leaf and/or shoot angles (Posada et al., 2012). Changes in petiole, leaf or shoot arrangement, such as leaf emergence pattern (Kikuzawa, 2003), might be an energetically cheap and efficient plastic response to reduce self-shading and may enhance light interception. An alternative explanation may be that the consequences of selection for long lifespan outweigh those of selection for an optimal light capture. Likewise, the study of Sterck et al. (2013) conducted on 15 tropical shrubs species at the seedling stage showed that the plastic variation of leaf area ratio (leaf area per unit plant mass) is likely to be favourable to carbon gain albeit unfavourable to light interception efficiency. Furthermore, plants can also achieve a low whole-plant compensation point by a low leaf compensation point (Sterck et al., 2013).

Concerning within-species variations along a light gradient, we observed that relative crown depth was shorter in low-light than in high-light conditions. This pattern corroborates previous studies conducted in tropical forest (Poorter and Werger, 1999; Sterck et al., 2001). We found that plastic variations in relative crown depth reflected variations in crown vertical growth rate rather than adjustment of leaf or branch longevity: shallow crowns of low-light trees predominantly result from slow crown vertical growth. We did not find evidence of a reduction in LLS or of accelerated branch shedding under deep shade, which would be the sign of self-shading or of an adaptive strategy of nutrient recycling. In contrast, we found stable or longer LLS in the shade. The growth rate of heliophilic species is known to be more sensitive to light constraints than that of shade-tolerant species as a by-product of specialization to high-resource habitats (Lortie and Aarssen, 1996; Laurans et al., 2012). Accordingly, we observed this response within four heliophilic species. A positive correlation between crown vertical growth and relative crown depth suggests that relative crown depth could be used as a surrogate of height growth rates of shade-intolerant species. King and Clark (2004) showed among three tropical tree species the strong correlation between leaved stem length and height growth rate. The emerging within-species pattern revealed in our study was consistent with Givnish’s assumption (shallower crowns expected under lower light) but the underlying mechanism, namely reduction in crown growth, does not support the interpretation in terms of self-shading reduction.

Are crowns of shade saplings more efficiently organized for light capture?

Whole-plant energy capture depends critically on the integration of foliage into an efficient canopy (Valladares and Niinemets, 2008), and thus the consideration of tree morphology variations among and within species might be essential for understanding light assimilation and growth strategies. We characterized the morphological plasticity of four species differing in light-niche position and breadth and looked at the extent to which a combination of crown structure and foliage adjustments may favour light capture in low-light conditions and may explain relative species performance along the light gradient.

Plasticity in crown slenderness

Plasticity in crown slenderness was significant for the two most light-demanding species, Q. rosea and V. michelii, and to a lesser extent for S. sp1. The reduction in crown slenderness in shade resulted from a reduction in crown vertical growth discussed above for V. michelii and S. sp1, and from adaptive adjustments in growth allocation between vertical and horizontal directions for Q. rosea (Table 5). Qualea rosea grew two times faster in the horizontal direction (first-order branches) than in the vertical direction (main axis) in the shade. Such a response of crown structure, which potentially increases light capture, has been described in shade-tolerant conifer species whose crown displays an umbrella form in shade (O’Connell and Kelty, 1994).

Plasticity in leaf dispersion

We observed decreasing branch and leaf spacing with decreasing light for Q. rosea, V. michelii and to a lesser extent for S. sp1, as a consequence of a reduction in internode length. Shorter branch spacing is likely to impact negatively light interception by increasing self-shading but at the same time it could reduce structural costs of foliage support (Pearcy et al., 2005). King et al. (1997) suggest that the close branch spacing of shaded saplings may be an expression of low allocation to stem growth in energetically constrained plants. It is noteworthy that no species, even among the most light-demanding, expressed shade-avoidance response traits such as an increase in internode length or in stem slenderness in shaded conditions.

Plasticity in total leaf area and ecological performances

Virola michelii was the most shade-intolerant and the most specialized species and the only species out of four to display a significant reduction in total leaf area (around 30 %) in shaded conditions. The number of first-order branches decreased in shade because of the concomitant reduction in leader stem growth rate and unchanged LLS. In addition, the number of leaves per branch was also reduced because of the diminution in leaf emission rate and concomitant steady LLS. Indeed, we observed a reduction in leaf cohort length on low and middle branches in low-light conditions but not on upper branches. This reduction in leaf number on the lower branches is likely to counter the reduction in stem and branch internode length and tier spacing under shade. The absence of significant plasticity in LLS for V. michelii is remarkable. It could result from the interplay in shaded conditions of the reduction in metabolic activity which enhances LLS and the occurrence of a negative carbon gain at the leaf scale which by accelerating leaf senescence may decrease LLS. Thus, the variation in the interplay of LLS and leaf emission rate on the main stem and on first-order branches may explain the decline of total leaf area in the shade observed for shade-intolerant species such as V. michelii. These species would not produce enough leaves or branches to compensate for their short leaf and branch lifespans.

The absence of a significant reduction in total leaf area is noteworthy in the case of Q. rosea as this species in the shade showed a significant reduction in crown depth and in leaf and branch emission rate. This pattern can be explained by the strong increase of LLS (from 23 to 37 months) in the shade, which may counter the reduction in leaf emission rate. Our observations at the branch level confirmed this interpretation as the length of leaf cohorts was constant along the light gradient for low and middle branches and even higher for upper branches in low-light conditions. The excess of leaf number supported by upper branches compared with lower branches may result from their higher leaf emission rate. Concerning the leader stem, the reduction in branch number under shade indicates that an increase of LLS did not outweigh the strong reduction in stem growth rate. The preferential growth investment in upper plagiotropic branches enables increased light capture and extends, in combination with the plasticity of LLS, the maintenance of efficient foliage in shaded conditions. This morphological response might contribute to a better survival of Q. rosea than V. michelii in low-light conditions and explain the lower degree of specialization of Q. rosea (Table 1).

The case of S. sp1 is interesting as this species closely resembled the most shade-tolerant species in some traits such as mean values of leaf mass per area, LLS and a low morphological plasticity while showing affinities with the less shade-tolerant species in some other traits such as mean growth rate. The only morphological response that we detected for this species was a release of apical control in the shade, which could cause the observed shift in crown slenderness. Total leaf area of S. sp1 was also constant along the light gradient, consistent with the maintenance of its crown length but in contrast to the strong decrease of stem growth rate. Here again, variation in LLS (from 31 to 56 months) is likely to counter variation in the leaf emission rate of branches. At the branch level, we did not observe any significant variation in leaf emission rate along the light gradient whatever the branch position on the stem. In this context one could expect an increase of total leaf area in shade. The decrease of unit leaf area in the shade (Table 4) might be one possible explanation for this constancy of total leaf area.

CONCLUSION

This study examined morphological diversity and plasticity in response to canopy openness of saplings of 14 tree species of French Guiana differing in light-niche optimum and breadth. The findings help to disentangle the adaptive response or strategies and constraint effects affecting crown shape of tropical saplings. Branching patterns are strongly controlled by physiological, biomechanical and environmental factors, as well as by genetic factors under phylogenetic constraints so their evolution is the result of reconciling these different design requirements (Seiwa et al., 2006). Integrating and scaling-up leaf-level dynamics to the shoot and crown level helps to interpret in adaptive terms inter- and intraspecific patterns of crown traits and to better understand species growth strategies. Although inter- and intraspecific variation of crown depth showed opposite patterns, there is no evidence to support the hypothesis of their adaptive value regarding enhancement of light capture efficiency in shade. This study confirms that Givnish’s hypothesis stating maximization of whole-plant carbon gain and the stress-tolerance hypothesis of shade-tolerance are not mutually exclusive: long LLS enables the maintenance of an extensive leaf area, which increases low-light carbon gain, while at the same time high leaf mass per area would increase tolerance to biotic stresses. Our findings provide evidence of a close link between LLS, relative crown depth and degree of shade tolerance. Consequently, relative crown depth might be a surrogate of low-light survival ability, and is easier to measure than LLS. Such a marker of shade tolerance has been applied in temperate sylviculture (Lorimer, 1983). At the intraspecific level, foresters also need simple surrogates of vigour to select trees with good future growth and survival prospects (Sterck et al., 2003). Relative crown depth may be useful in making qualitative assessments of growth and assessing the vigour of individual saplings.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Figure S1: links between leaf-, branch- and crown-level traits contributing to light capture in tropical saplings. Figure S2: light-niche optimum (POP-RESP) and light-niche breadth (Degree of specialization) of the 14 tropical tree species involved in the present study.