• Here, we tested two hypotheses: shading increases light interception efficiency (LIE) of broadleaved tree seedlings, and shade-tolerant species exhibit larger LIEs than do shade-intolerant ones. The impact of seedling size was taken into account to detect potential size-independent effects on LIE. LIE was defined as the ratio of mean light intercepted by leaves to light intercepted by a horizontal surface of equal area.
• Seedlings from five species differing in shade tolerance (Acer saccharum, Betula alleghaniensis, A. pseudoplatanus, B. pendula, Fagus sylvatica) were grown under neutral shading nets providing 36, 16 and 4% of external irradiance. Seedlings (1- and 2-year-old) were three-dimensionally digitized, allowing calculation of LIE.
• Shading induced dramatic reduction in total leaf area, which was lowest in shade-tolerant species in all irradiance regimes. Irradiance reduced LIE through increasing leaf overlap with increasing leaf area. There was very little evidence of significant size-independent plasticity of LIE.
• No relationship was found between the known shade tolerance of species and LIE at equivalent size and irradiance.
To acquire the energy required for the numerous processes involved in the conversion of atmospheric CO2 into biomass, plants need to capture solar radiation within their surrounding environment. In a forest understory, where light is often highly variable in time and in space (Runkle, 1982; Canham et al., 1990; Gendron et al., 2001), growth and survival of regenerating seedlings are closely linked to their capacity to intercept light efficiently under contrasting light environments, and this capacity may be critical to seedling survival.
Differences in irradiance received during development resulted in a large intraspecific diversity in LIE for single branches or whole crowns (phenotypic plasticity) (Planchais & Sinoquet, 1998; Valladares & Pearcy, 1998; Farque et al., 2001; Fleck et al., 2003; Gàlvez & Pearcy, 2003). Under low irradiance regimes, LIE is frequently increased because of smaller leaf inclination angles (i.e. horizontal leaves) or a smaller number of leaf layers or petiole twisting, the two latter processes limiting the detrimental effect of self-shading among leaves. Conversely, LIE is decreased under higher irradiance regimes because of denser crowns and larger leaf area that increase leaf overlap and self-shading. However, this decrease in LIE with higher irradiance regimes is often counterbalanced by a deeper light penetration within the crown and results in reduced interception of damaging high and direct sunlight radiations.
Here, we examined the variation in leaf organization and integrated crown light interception of five different broadleaved species: (i) two North American species, the shade-tolerant Acer saccharum and the moderately shade-tolerant Betula alleghaniensis; and (ii) three western European species, the shade-tolerant Fagus sylvatica, the moderately shade-tolerant A. pseudoplatanus, and the pioneer and shade intolerant B. pendula. We used direct measurements and spatially explicit three-dimensional model simulations to compute light interception. For this purpose, we digitized the crown of 1- and 2-year-old seedlings grown in a nursery under three contrasting light regimes (36, 16 and 4% of full sunlight, provided by neutral shading nets) to assess the precise organization of leaves within the crown volume. Then, using reconstructed three-dimensional images of digitized seedlings in the plant functional and structural model VegeSTAR (Adam et al., 2004), silhouette to total leaf area ratio was estimated for single directions (STARΩ) or integrated at crown level over the whole sky (STARSky) and then used for estimating whole-plant LIE.
This study had three main objectives: (i) to assess the plasticity of leaf organization and resulting crown LIE in relation to contrasting irradiance microclimates and seedling age (ontogenic processes); (ii) to compare species for the degree of plasticity they show in LIE; and (iii) to check whether shade-tolerant species differ from shade-intolerant ones with respect to LIE.
Materials and Methods
Plant material and experimental design
Two North American broadleaved tree species, Acer saccharum Marsh (sugar maple: AS) and Betula alleghaniensis Britton (yellow birch: BA), and three European ones, Acer pseudoplatanus L. (sycamore: AP), Betula pendula Roth (European white birch: BP) and Fagus sylvatica L. (European beech: FS) were used. Germinated seedlings were transplanted into 10 l pots filled with a mixture of peat and sand (1/3 v/v) and placed in a nursery at Champenoux (48°44′ N, 6°14′ E) near Nancy, France. We used three aluminized neutral shading nets (OLS screens, Ludvig Svensson, Kinna, Sweden) which provided three contrasting irradiance regimes [high light (HL), 36% of the corresponding external irradiance; intermediate light (IL), 16%; and low light (LL), 4%; further discussed later]. Seedlings were placed at least 30 cm from each other to limit mutual shading. They were fertilized twice during the growing season (during May and at the beginning of August) with 40 g (4 g l−1 substrate) of a slow-release fertilizer (Nutricote® 100+, 13/13/13 N/P/K+ oligo-elements). Water was provided every day to field capacity using drip irrigation controlled by a time switch. Three series of 30 seedlings per species and treatment were planted during May 2000, 2001 and 2002 under the same shading nets so as to obtain simultaneously 1- and 2-year-old seedlings for measurements. During the summer, one to six healthy individuals were selected for digitization. Table 1 provides details of seedling dispatching among years, ages and species.
Table 1. Number of digitized saplings according to species, light regime, age and installation year
Measurement of irradiance microclimate below shading nets
Two linear PAR sensors (amorphous silicium sensors, Solems, Palaiseau, France) recorded continuously (pace: 1 s, averaged values over 30 mn stored in a CR10X data logger, Campbell Scientific, Logan, UT, USA) photon flux density in the PAR (PPFD) below each net, while a similar sensor was used for incident irradiance outside the net. Sensors were calibrated every year with a Li-Cor reference quantum sensor (Li-190, Li-Cor Biosciences, Lincoln, NE, USA). A Delta-T BF2 sensor (Delta T, Cambridge, UK) was used to record diffuse and direct PPFD. Transmittance in the PAR below the shading nets was calculated as the ratio of the sum of PPFD below and above the nets. Resulting levels were as follows: HL, 36%, IL, 16%, LL, 4%.
During August 2000, a linear PAR sensor Li-Cor Li-191 (Li-Cor Biosciences) and a diffuse-direct sensor Delta-T BFS (Delta T) were used during two clear days (17 and 23/08/00) between 10:00 and 15:00 h local time, to assess diffuse and direct irradiance above and below the nets. Transmittances for direct and diffuse irradiance were expressed as a fraction of the incident PPFD: HL, 25.5% direct and 121% diffuse incident PPFD; IL, 8.8% direct and 60.0% diffuse PPFD; LL, 1.0% direct and 29.7% diffuse PPFD. The greater than 100% diffuse PPFD estimated below HL shading nets was attributable to redistribution of direct PPFD by multiple scattering inside the aluminized nets. This shows that the nets altered significantly the ratio diffuse/direct irradiance, the latter component being minor under the nets. Spectral composition of the light was not modified by the nets, and the red/far red ratio remained unaffected (data not shown).
A total of 136 seedlings with two ages, four irradiance regimes and five species (i.e. between one and six replicates per individual treatment) were digitized during the 2000, 2001 and 2002 growing seasons (Table 1). Spatial distribution and geometry of leaves within the crown volumes was recorded using a Fastrak electromagnetic 3D apparatus (Polhemus Inc., Colchester, VT, USA) as described by Sinoquet et al. (1998). This apparatus associated with the 3A computer software (Adam et al., 1999) allowed a precise positioning of each leaf. Position and orientation as given by the three Euler angles (leaf midrib elevation, θ; leaf inclination around midrib, Φ; and leaf midrib azimuth, Ψ) were recorded for each individual leaf assumed to be flat. This method, first tested by Sinoquet and Rivet (1997), was recently applied to describe the canopy structure of a range of plant species (Sinoquet et al., 1998; Rakocevic et al., 2000; Falster & Westoby, 2003; Sonohat et al., 2006). To prevent canopy movements induced by wind and interferences within the electromagnetic measurement field (Sinoquet & Rivet, 1997), digitization was performed inside a small wood cabin that contained no metallic elements. One measurement at the lamina-petiole junction described the leaf spatial coordinates and Euler angles since the sensor was held along the midrib and parallel to the leaf blade. For the North American species, leaf length was estimated by recording spatial coordinates of lamina extremity (error < 1 mm, Sinoquet et al., 1998), whereas for the European species, leaf length and width were measured to the nearest mm with a ruler. Time required for digitization was from 10 min in the case of a 20-cm-high A. saccharum to 2 d for a 190-cm-high B. alleghaniensis.
Canopy reconstruction of digitized individuals and light interception simulations
The geographically oriented x, y and z coordinates, and the three Euler angles recorded for each leaf after seedling digitization were used to reconstruct the crown of each individual using the three-dimensional explicit model VegeSTAR 3.02 (Adam et al., 2004; Fig. 1). For each species, the specific relationship between leaf area (a), leaf length (l) and leaf width (w) was computed from a sample of leaves (c. 30 leaves per species) from the seedlings. The resulting relationships (r2 ≥ 0.95) were as follows: a = 0.75l2, A. saccharum; a = 0.34l2, B. alleghaniensis; a = 0.61lw, B. pendula; a = 0.57lw, A. pseudoplatanus; a = 0.73lw, F. sylvatica. These relationships were used to compute total leaf area per seedling. For each species, leaf surfaces were visualized in the model by a polygonal leaf prototype that mimicked real leaf shape and obeyed the relationship between leaf area and dimensions.
To assess LIE, we used the ratio STAR (Carter & Smith, 1985; Oker-Blom & Smolander, 1988). The use of STAR to estimate LIE is based on several assumptions: (i) light scattering is neglected, i.e. leaf absorptance equals 1; and (ii) light interception by branches and twigs is negligible. Directional STAR was computed with the VegeSTAR 3.02 software (Adam et al., 2004; Sinoquet et al., 2005) as the ratio of projected leaf area in a given direction, as viewed by an orthographic camera (i.e. a camera with parallel beams), to the total leaf area. We used 46 directions pointing from the sky hemisphere to the three-dimensional model, with approximately equal adjacent solid angles (c. 2π/46 steradians) and penta- and hexagonal facets (Den Dulk, 1989). The distribution is further referred to as ‘turtle sky’. Each direction Ω was defined by elevation (ΩEl) and azimuth (ΩAz) angles.
where Φ, θ and Ψ are the leaf rolling, elevation and azimuth angles, respectively.
STAR and the leaf inclination component Li were integrated over the whole sky hemisphere (STARSky and LiSky). For this purpose, the 46 values of STAR and Li were weighted using an estimate of the contribution of cumulated irradiance incoming from each turtle sky facet below the shading nets during the growing season (1 May to 30 September):
LoSky, leaf overlap component integrated over sky hemisphere, was calculated from LiSky and STARSky as follows:
LoSky = STARSky/LiSky(Eqn 6)
LIE as defined in the introduction was calculated as follows:
According to this definition, LIE is equivalent to STARSky (weighted average of directional STAR) scaled so that a horizontal leaf has a LIE value equal to one under the considered light regime. Multipliers for converting STARSky to LIE are 1.91, 1.93 and 1.96 for HL, IL and LL regimes, respectively.
Cumulated irradiance incoming from each turtle sky facet was calculated using Gap Light Analyser (GLA) software (Frazer et al., 1999). The sky hemisphere was divided into 4050 sectors (90 azimuth × 45 zenith angles) and incident irradiance computed for each sector. We used a virtual open sky photograph (i.e. with white pixels only) and attributed each GLA sector to a turtle sky facet. In the open, the fraction direct irradiance (%DirectOpen) was 50% (integrated measurements from 16 June to 30 September 2002). The fraction of direct irradiance below shading nets was derived from %DirectOpen and the measured relative direct PPFD below each net. Total diffuse radiation was computed from the measured relative diffuse PPFD of each net. The spatial distribution of diffuse irradiance was set to Uniform OverCast (UOC) because large redistribution of light beams occurs below the nets.
All statistical analyses were performed with the statistical software R version 2.0.1 (R Development Core Team, 2004). Linear models were fitted for testing species, age and irradiance regime effects on response variables. When required, an additional covariate was included in the model for scaling the response variable to plant size. The log-transformed total leaf area At appeared to be the most suitable covariate as most size-dependent variables responded linearly to it (see the Results section). Our general approach was first to fit the main model with every interaction between factors and possibly covariate, up to the highest order, of the form:
Rijkr = µ + (ai + bj + ck + dij + eik + fjk + gijk) + ɛijkr(Eqn 8)
or, when the covariate was present:
(indices i, j and k, first, second and third factor levels, respectively; r, replicates; µ, intercept; ai, bj and ck, main effects on intercept; dij, eik and fjk, second-order interactions; gijk, third-order interaction effects on intercept, α, slope; effects on slope; Rijkr, response; Vijkr, covariate; ɛijkr, error term).
Then a cascade of nested models were fitted by setting to zero or equalizing the parameters that were not significantly different (P < 0.05) from zero or not significantly different among each other. Highest-order interaction parameters were checked first. At the end of the procedure, the most parsimonious (i.e. with the lowest number of parameters) model that was not significantly different (F-test between main and submodel) from the main model was retained. At this stage, and if required, multiple comparison tests using the multivariate t distributions of the parameters were performed on parameters of interest with the multcomp R-package (Bretz et al., 2004). Only the results of this final model are detailed in this paper.
Normality and homoscedasticity were checked graphically by plotting residuals against predicted and by normal quantile to quantile plots.
Total leaf area
Total leaf area of saplings (At) differed significantly among irradiance regimes and among species at all ages (Table 2). The general pattern was an increase in At with increasing irradiance. During the first year A. pseudoplatanus and B. pendula developed largest At as compared with the other species under all irradiance regimes. During the second year, B. alleghaniensis had a large growth and as a result the largest leaf area, followed by B. pendula. After two growing seasons, the two Acer species were intermediate in almost all treatments. Meanwhile, F. sylvatica had dramatically low values of At as compared with any other species and exhibited an eightfold difference with B. alleghaniensis under high irradiance (HL, Fig. 2).
Table 2. Results of a two-way anova on loge(total leaf area) and light interception efficiency (LIE) of 1- and 2-year-old seedlings as a function of irradiance regime (36, 16 and 4% of external irradiance) and species
loge(total leaf area)
P > F
P > F
P > F
Leaf inclination distribution
In order to document the leaf inclination component of STARSky, we calculated the distribution of leaf inclination, defined as the angle between the leaf normal and the vertical, and calculated as arcos(| cos(Φ)cos(θ) |), relative to leaf area (and not to leaf number to limit the influence of small nonrepresentative leaves) in 2-year-old seedlings (Fig. 3). All species but B. pendula showed a shift toward less horizontal leaves with increasing irradiance regimes. B. pendula showed a more evenly distributed leaf inclination than other species with no clear evidence of a shift attributable to irradiance regime.
Light interception efficiency
Light interception efficiency differed significantly among irradiance regimes and species during both the first and second years, according to a two-way anova (Table 2). However patterns of variation were much less straightforward than for At (Fig. 2). Nevertheless a general decrease in LIE was detected with increasing irradiance both during the first and during the second year. Moreover, second-year values were smaller than corresponding first-year values. Differences among species were difficult to detect, as the species ranking depended on age and irradiance microclimate. This complex pattern was attributable to pronounced age- and size-related effects that interacted with irradiance-induced plasticity and with species effects.
Light interception efficiency decreased severely with increasing At (Fig. 4), and effects of irradiance regime and age on LIE might be confounded with ontogenic effects mediated by size (i.e. by At). We therefore introduced loge(At) as a covariable into the main linear model, according to Eqn 9 (n = 137, d.f. = 78, r2 = 0.849). The most parsimonious model (n = 137, d.f. = 129, r2 = 0.703) that was not significantly different from the main model (F = 1.484, d.f. = 51 and 78, P = 0.057) led to the relationships drawn in Fig. 4, and to the following conclusions: (i) no irradiance regime effect could be detected independently from that mediated by total leaf area; (ii) no age effect could be detected except for slope of 1-year-old B. pendula (P = 0.024); (iii) the slopes were not significantly different among species, except for the slope of the 1-year-old B. pendula (P = 0.024), that is, the size-induced plasticity of LIE was of similar amplitude in all species except 1-year-old B. pendula; (iv) there were significant differences in intercepts (P < 0.001) – multiple comparison tests produced the following ranking of species: FSa < ASb < BAb < BP (second year)c < APc (species sharing same letter are not significantly different). This means that F. sylvatica seedlings showed, after removal of the size effect, a smaller LIE in all treatments and all ages. The largest LIE was recorded in A. pseudoplatanus and B. pedula.
Leaf inclination component (LiSky)
Comparison of models according to Eqn 9 (all effects and interactions plus covariate) and Eqn 8 (all effects and interactions without covariate) revealed no significant effect of At on LiSky (F = 1.00, d.f. = 29 and 78, P = 0.479). This leads to the following conclusions: (i) the effect of leaf inclination on STAR remained constant with At; (ii) despite a very narrow range of LiSky (0.496–0.523), there were significant main effects of species, irradiance and age and significant interactions, age × irradiance and age × species, in the model according to Eqn 4 (Fig. 5); multiple comparison tests on irradiance led to LL < IL = HL, and on species to BPa < ASa < BAab < APab < FSb (treatments sharing the same letters are not significantly different).
Leaf overlap component (LoSky)
Contrary to LiSky, LoSky decreased linearly with log(At) (Fig. 6) in all species. A model according to Eqn 9 (n = 137, d.f. = 78, r2 = 0.835) was fitted to the data. The most parsimonious model (n = 137, d.f. = 129, r2 = 0.680) that was not significantly different from the main model (F = 1.331, d.f. = 51 and 78, P = 0.078) led to the linear regression lines drawn in Fig. 6 and to the following conclusions: (i) no irradiance regime effect could be detected independently from that mediated by total leaf area; (ii) no age effect except for slope of 1-year-old B. pendula (P = 0.028); (iii) the slopes were not significantly different between species, except for slope of 1-year-old B. pendula (P = 0.028), that is, the size-induced plasticity of LoSky was of similar amplitude in all species except 1-year-old B. pendula; (iv) significant differences in intercepts – multiple comparison tests produced the following ranking of species: FSa < ASb = BAb < BP (2nd year)c = APc (species sharing the same letter are not significantly different) (P < 0.001).
Directional effects on STAR
The results of the simulation presented above may be modified because of irradiance geometry, which changes dramatically from open spaces (similar contributions of direct and diffuse radiation) to deep shade below canopies (diffuse radiation dominant except during light flecks). To document this question, STAR was computed as a function of the elevation of incident irradiance (directional STAR, Fig. 7). In all species, STAR generally increased with elevation. Irradiance microclimate induced some changes, and STAR from highest elevations was smaller in plants grown under high irradiance regimes except in 2-year-old B. pendula.
A simple framework for analysing LIE variations
In this study, we used Pearcy's LIE decomposition (Pearcy & Yang, 1996) in order to distinguish light interception ability attributable to foliage orientation – the more horizontal leaves, the more efficient foliage – and foliage overlapping. Alternative frameworks relating to the theory of radiative transfer have been proposed to assess the effect of structural parameters on light interception properties (e.g. space occupation, leaf area density, foliage dispersion; Planchais & Sinoquet, 1998). However, the application of the radiative transfer theory to isolated plants depends on the definition of a canopy envelope filled with the vegetation elements (Norman & Welles, 1983; Cescatti, 1997). Recent studies have reported how subjective the definition of such canopy envelopes is, because the canopy volume depends on the shape of the envelope (e.g. bounding box vs convex envelope vs parametric shapes; Boudon, 2004) and also because of the fractal nature of canopies making canopy volume change with measurement scale (Phattaralerphong & Sinoquet, 2005). This is especially crucial in small plants like seedlings with a small number of leaves (Farque et al., 2001) because it is unclear if air space between leaves belongs to the canopy volume or not. As parameters like canopy porosity, leaf area density and, consequently, clumping depend on how canopy volume is defined, they have a similar uncertainty as canopy volume (Sinoquet et al., 2005). This is the reason that we have used a simple but robust framework – as previously proposed by Pearcy and colleagues – rather than an apparently more detailed one, but which shows some drawbacks (especially the scale-dependence of the canopy envelope).
Comparison between YPLANT and VegeSTAR
YPLANT (Pearcy & Yang, 1996) is a widely used model and, before further discussion, it may be useful to compare YPLANT with VegeSTAR. VegeSTAR only computes STAR, which is the ratio of projected leaf area to total leaf area (called ED in YPLANT). STARSky is derived after summing directional values over the sky vault, weighed by cumulative irradiance. By contrast, YPLANT also computes light absorption and interception efficiency, called Ea and Ei, respectively. Absorption computations involve leaf reflection and transmission processes, which are treated in a simplistic manner – in particular, the directional redistribution of scattered light is neglected. Interception efficiency assumes leaves as black bodies, i.e. leaf absorbance is 1. In conclusion, similar variables between VegeSTAR and YPLANT are directional STAR (in VegeSTAR) and ED (display efficiency in YPLANT). All other parameters in YPLANT such as Ei (LIE in this study), except for those implying absorption, can be calculated from the VegeSTAR output.
Total leaf area
The different irradiance regimes imposed on the seedlings from five European and North American broadleaved tree species induced large differences in growth. There was at least a sevenfold difference between minimum and maximum mean total leaf area (At) at the diverse ages vs irradiance regime combinations. This value needs be compared with the range of values affecting LIE: the variation encompassed only a maximal range of 1.3-fold. Therefore, total leaf area developed by the seedlings accounts for most of the interspecific diversity of total light interception. LIE plays only a minor role in this variability.
Species ranking for total leaf area remained stable among the different irradiance regimes, but changed with age. During the second year of growth, the ranking from smaller to larger total leaf area overlapped with the ranking for shade tolerance. The shade-tolerant F. sylvatica had, in all cases, the lowest total leaf area. A. pseudoplatanus and A. saccharum, with intermediate tolerance, ranked intermediate and the two more shade-intolerant birches (B. pendula and B. alleghaniensis) had the largest leaf areas. Similar findings have been reported by Kitajima (1994), Walters and Reich (1999) and Poorter (1999), but this is not a general rule (Montgomery & Chazdon, 2002). This large difference in sapling size (estimated from leaf area) among species is probably one of the clues to shade tolerance/intolerance, as discussed later.
Interspecific variability of LIE
Static observations and descriptions of fine-scale tree traits generally show a large diversity in crown morphology and architecture among tree species (Halléet al., 1978; Sterck, 1997; Millet et al., 1998; Pearcy et al., 2004; Claveau et al., 2005). Indeed, our seedlings revealed a large diversity in growth and branching patterns, with a sympodial growth in beech and a monopodial growth in the other species, with differences in ramification, leaf sizes and leaf inclination. But with respect to LIE, our results are in line with some recent findings (Valladares et al., 2002; Pearcy et al., 2004) that revealed only small differences when comparing species under similar irradiance environments and similar sizes/ages. This suggests a strong interspecific convergence in crown light capture capability (Poorter & Werger, 1999; Valladares et al., 2002), possibly attributable to constraints imposed by the other functions of the crown, such as water transport and mechanical resistance (Pearcy et al., 2005).
Irradiance-induced vs ontogenic plasticity of LIE
Light interception efficiency was found to vary dramatically in all species with changes in size (leaf area being an estimator of size); that is, there was a large ontogenic plasticity in LIE. A large fraction of the effect of shading on total light interception and LIE was therefore induced by delayed growth in shade vs high light grown saplings (Coleman et al., 1994). Indeed, as leaf area accumulated during growth, leaf overlap severely depressed LIE, as was observed by Farque et al. (2001) in oaks. By contrast, irradiance-induced changes in leaf inclination tended to have a small influence on LIE, owing to the spatial integration over a sky hemisphere where irradiance distribution is essentially homogeneous after light redistribution below the shading nets. This result contradicts some earlier studies that reported a large impact of leaf inclination in the modulation of LIE (Takenaka et al., 2001; Valladares et al., 2002). However, these latter findings referred to particular species such as understory fruit plants, steep-leaved herbs, palms or lianas, where individual architecture does not experience equivalent mechanical and functional constraints to those of long-lived and potentially tall forest trees and to understory grown plants undergoing much more heterogeneous light distribution than found under our shading nets. In these examples, the plants were also small, with few leaves, so that leaf overlap was of less influence on LIE than leaf inclination. Nevertheless, saplings with higher leaf area (i.e. 2-year-old saplings and saplings grown under higher irradiance) had more leaves with larger inclination angles, which might increase interception of irradiance from more horizontal directions. We found that this effect remained small in all conditions. Moreover, the largest differences attributable to irradiance regimes occurred at the highest angular elevations. Generally the leaf overlapping component of STAR decreased (i.e. there was more leaf overlap), while the leaf inclination component increased with increasing elevation of incident irradiance. For plants grown under higher irradiance, the smaller STAR at directions close to vertical is again attributable to leaf overlap. It is not compensated for in the horizontal directions by a shift in leaf inclination distribution to more vertical inclinations because, despite this shift, there was still a majority of leaves whose inclination is closer to horizontal than to vertical. A noticeable exception was B. pendula, the most shade-intolerant among the studied species, where the STAR gradient from horizontal to vertical directions was smooth and not influenced by irradiance regime. This is attributable to the conical crown shape of this species (Fig. 1), combined with its even leaf inclination distribution and its ability to shed its innermost shaded leaves: its foliage is distributed like a carapace, allowing light interception to be equally efficient whatever the direction of incident light, which might give an advantage in open sites.
Light interception and seedling functional ecology
Similar conclusions describing little or no interspecific differences in light interception abilities under similar environments (Poorter & Werger, 1999; Valladares et al., 2002; Pearcy et al., 2004) do not rule out the supposed importance of LIE in the diversity of species functional ecology or shade tolerance. Indeed, small differences between species in instantaneous or daily light capture may lead to rather large differences in final carbon gain over the whole growing season. Here, substantial differences between the five species (in LIE or LoSky) demonstrated that slight architectural and phyllotaxic species-specific traits were maintained even if the whole-crown LIE was similarly affected among species by dimensional constraints. However, the ranking of species for LIE differed largely from that for shade tolerance. Noticeably, the most shade-tolerant species, F. sylvatica, had the lowest LIE among the five species studied.
LIE was larger in individuals grown under low than under high irradiance. This was mainly an effect of irradiance-induced differences in seedling sizes rather than a light-driven acclimation of architectural and phyllotaxic traits. As demonstrated in several earlier studies, individual size greatly affects morphological and allocational patterns, and apparent plasticity in response to environmental factors such as light, nutrients or water may be at least partly attributed to such ontegenetic shifts during plant ontogeny (Huber & Stuefer, 1997; McConnaughay & Coleman, 1999; Müller et al., 2000). Specifically, tree size modulates the ratio of photosynthetic to nonphotosynthetic tissues (Messier & Nikinmaa, 2000; Naumburg et al., 2001; Claveau et al., 2002; Delagrange et al., 2004). This is of particular interest with regard to seedling early development in various understory environments. Decreasing LIE with increasing individual size and leaf area tends to support the hypothesis that at the establishment stage, seedlings are more efficient in light capture, especially in elevations with potentially high returns in direct light (i.e. vertical directions) than at further developmental stages. Then, as seedlings are growing and accumulating biomass and leaves, leaf overlap increases and the ratio of photosynthetic to nonphotosynthetic tissues declines, so that the plants may require more light for achieving a positive carbon balance or, under constant light, reach a maximum size over which carbon balance becomes negative, leading to the death of the tree (Messier et al., 1999; Messier & Nikinmaa, 2000). Therefore, long-term survival under shade can be achieved either by delaying the time when maximum size is reached through low growth or by allowing a greater maximum size through minimizing carbon costs of leaf display and carbon assimilation or both.
Thus, species-specific patterns of carbon allocation under contrasting light environments might be critical for their LIE and resulting functional ecology. Species allocating carbon preferentially to above-ground growth even under shaded environments will reduce their LIE, since they are increasing their size and leaf area, whereas species maintaining lower growth rates under shaded conditions might increase their survival by maintaining a more efficient leaf display with a low carbon cost. These specific differences may help to explain the larger shade tolerance of F. sylvatica, A. pseudoplatanus and A. saccharum compared with Betula species in the forest understory. Indeed, Acer species are able to control apical and especially lateral meristem development when suppressed in order to maintain very slow growth rates under very low light conditions (Sipe & Bazzaz, 1994; Goulet et al., 2000; Delagrange et al., 2004). F. sylvatica controls its development through polycyclism: in shaded environments this species exhibits only a single flush in spring, then buds enter dormancy and no leaf production occurs later on. By contrast, Betula species continuously produce new leaves while possibly shedding the older, shaded ones: species that continuously allocate biomass to height and lateral growth of crown, even in shaded environments, are likely to increase carbon costs and self-shading dramatically (Kikuzawa, 2003).
Furthermore, minimizing the cost of leaf display through architectural traits could give an advantage under shade. The studied species differed greatly according to their developmental patterns which, as pointed out by Beaudet and Messier (1998), may act as important constraints to morphological acclimation to the light regime. For instance, Acer species which have long petioles and large leaves may improve their leaf display efficiency at lower cost (Beaudet & Messier, 1998; Takenaka et al., 2001) than highly branched Fagus species with small leaves and short petioles and where one possible way to decrease leaf overlap would be through increasing internode length but it could rapidly have a detrimental effect on carbon balance (Pearcy et al., 2005). Crown shape may also be important for light capture as horizontally spread crowns such as those of Fagus are expected to have a higher efficiency for intercepting light incoming from vertical directions than erect crowns such as those of Acer or, to a lesser extent, Betula. But, as Fagus has the lowest LIE among the species studied, this trait cannot alone account for LIE interspecific variability. Obviously, LIE and its leaf overlap component are the outcome of a complex combination of architectural traits constrained by developmental pattern and associated carbon costs; more research, especially on carbon costs and gains associated to architectural traits, is needed to understand shade tolerance in trees better.
Finally, another essential factor associated with LIE and possibly linked to species functional ecology is the potential partitioning between direct and diffuse light capture among species. Although it is difficult to draw general conclusions, Fig. 7 clearly indicates that some species are able to adjust their diffuse vs direct light capture. Generally, 1-year-old seedlings preferentially intercept light from vertical directions, while 2-year-old seedlings do it mainly from all directions. Similarly, except for the shade-intolerant B. pendula, efficiency of interception in vertical directions increases with decreasing light regimes. Several studies have reported the ability of some tree species seedlings to preferentially intercept direct (Muraoka et al., 2003) or diffuse light (Ackerly & Bazzaz, 1995; Valladares et al., 2002). However, to estimate the advantage of enhancing diffuse light compared with direct light interception (or the reverse) in a given light environment and conclude on its impact on the survival of those species in forest understory, more research is needed: (i) to develop long-term light capture models based on realistic and highly variable cloudy/sunny conditions; and (ii) to improve our understanding of sun-fleck use (i.e. photosynthetic apparatus activation and de-activation).
S. Delagrange was supported for this research by a PhD grant from the Université du Québec à Montréal (UQAM) and by a bilateral international agreement with the Université Henri Poincaré, Nancy (thanks to Prof. Pierre Dizengremel, UHP Nancy, for his support). The help of Jean Marie Gioria in growing the plants and managing the experimental design at INRA Nancy, and that of Daniel Himbert and Delphine Retzinger in digitizing the saplings, is gratefully acknowledged. The project was partly financed by a research grant from Région Lorraine, and from GIP ECOFOR, Paris.