Leaf Structural and Photosynthetic Characteristics, and Biomass Allocation to Foliage in Relation to Foliar Nitrogen Content and Tree Size in Three Betula Species

Young trees 0´03±1´7 m high of three coexisting Betula species were investigated in four sites of varying soil fertility, but all in full daylight, to separate nutrient and plant size controls on leaf dry mass per unit area (M A), light-saturated foliar photosynthetic electron transport rate (J) and the fraction of plant biomass in foliage (F L). Because the site effect was generally non-signi®cant in the analyses of variance with foliar nitrogen content per unit dry mass (N M) as a covariate, N M was used as an explaining variable of leaf structural and physiological characteristics. Average leaf area (S) and dry mass per leaf scaled positively with N M and total tree height (H) in all species. Leaf dry mass per unit area also increased with increasing H, but decreased with increasing N M , whereas the effects were species-speci®c. Increases in plant size led to a lower and increases in N M to a greater F L and total plant foliar area per unit plant biomass (LAR). Thus, the self-shading probably increased with increasing N M and decreased with increasing H. Nevertheless, the whole-plant average M A , as well as M A values of topmost fully exposed leaves, correlated with N M and H in a similar manner, indicating that scaling of M A with N M and H did not necessarily result from the modi®ed degree of within-plant shading. The rate of photosynthetic electron transport per unit dry mass (J M) scaled positively with N M , but decreased with increasing H and M A. Thus, increases in M A with tree height and decreasing nitrogen content not only resulted in a lower plant foliar area (LAR = F L /M A), but also led to lower physiological activity of unit foliar biomass. The leaf parameters (J M , N M and M A) varied threefold, but the whole-plant characteristic F L varied 20-fold and LAR 30-fold, indicating that the biomass allocation was more plastically adjusted to different plant internal nitrogen contents and to tree height than the foliar variables. Our results demonstrate that: (1) tree height and N M may independently control foliar structure and physiology, and have an even greater impact on biomass allocation; and (2) the modi®ed within-plant light availabilities alone do not explain the observed patterns. Although there were interspeci®c differences with respect to the statistical signi®cance of the relationships, all species …


INTRODUCTION
Leaf dry mass per unit area (M A ) and the fractional investment of the biomass in foliage (F L ) are important plant parameters which are strongly linked to plant growth rates (Evans 1972;Poorter and Remkes, 1990;Ko Èrner, 1991;Cornelissen et al., 1996Cornelissen et al., , 1998)).The total leaf area per unit plant biomass (LAR) is given by F L /M A , and, thus, both a large F L and a small M A allow formation of an extensive foliar area, thereby enhancing plant potential for light interception and photosynthesis.Therefore, plant growth rates often vary directly with LAR (Ingestad and McDonald, 1989;Poorter 1990;Cornelissen et al., 1996Cornelissen et al., , 1998;;Wang et al., 1998).
Along gradients of nutrient availability, the biomass requirements for light harvesting con¯ict with those of nutrient uptake and, accordingly, the signi®cance of a large foliar area is greater at high nutrient availabilities and less at low nutrient availabilities.To cope with a broad variability in soil conditions, plants often possess a large capacity for plastic modi®cation of biomass allocation and total leaf area.Increased nitrogen supply invariably leads to a larger LAR (Karlsson and Nordell, 1987;Ku Èppers et al., 1988;Thompson et al., 1988;Ingestad and McDonald, 1989;Bowler and Press, 1996;Wang et al., 1998).Yet, it is still unclear to what extent these changes result from allocation (F L ) and from alterations in leaf structure (M A ).There are consistent positive relationships between F L and nutrient availability (Ingestad and Lund, 1979;Karlsson and Nordell, 1987;Ingestad and McDonald, 1989;Wang et al., 1998;Poorter and Nagel, 2000), but the evidence of nutrient availability effects on M A is contrasting.Strong interspeci®c negative correlations between M A and foliage nitrogen content per unit dry mass (Turner, 1994;Reich et al., 1995;Niinemets, 1999) are often found, but the relationships are weaker in other studies (Reich and Walters, 1994).In addition, increases in nitrogen availability have resulted in a lower whole-plant average M A (Jurik et al., 1982;Waring et al., 1985;Karlsson and Nordell, 1987;Hirose, 1988;Thompson et al., 1988;Latham, 1992;Bowler and Press, 1996;Grubb et al., 1996;Wang et al., 1998;Meziane and Shipley, 1999) but, in some cases, M A is invariable (Karlsson and Nordell, 1989;Kolb et al., 1990;Fredeen et al., 1991;Latham, 1992;Ackerly and Bazzaz, 1995;Grubb et al., 1996;Sims et al., 1998b;Rosati et al., 2000) or even increases (Latham 1992;Sims et al., 1998a) with increasing nitrogen availability.
Apart from the possible nutrient effects, M A scales positively with light availability (Latham, 1992;Ackerly and Bazzaz, 1995;Niinemets, 1998;Niinemets and Kull, 1998;Rosati et al., 2000).Given that the increase in foliar area as the result of improved nutrition inevitably leads to a greater within-plant shading (Ackerly and Bazzaz, 1995), a decrease in M A in plants at higher nutrient availability may be a consequence of lower quantum ¯ux densities within these plants.In addition, M A is also positively related to plant size (Steele et al., 1989;Niinemets, 1997a;Niinemets et al., 1999).Because the plants at higher nitrogen availability are generally larger, plant size effects on M A may level off or over-rule the in¯uences of enhanced shading, providing an explanation of insigni®cant or positive effects of improved N nutrition on M A .
We studied biomass allocation and leaf morphology in young trees in four sites of contrasting fertility to: (1) determine the extent to which leaf nitrogen availability controls M A , foliar photosynthetic characteristics and the fraction of plant biomass in foliage (F L ); and (2) understand whether these effects can be ascribed to modi®cations in intra-plant shading and/or differences in plant size.We also analysed (3) foliar construction cost in terms of unit glucose investment for a unit foliar biomass production (Vertregt and Penning de Vries, 1987;Poorter, 1994) to determine whether the carbon cost of leaves varies as a function of foliar nitrogen.We expected that in less fertile habitats, where carbon availability limits growth to a lesser extent than in more fertile patches, the content of carbon-rich protective chemicals such as lignin and condensed phenolics increases (Waring et al., 1985;Chapin, 1989;Chapin et al., 1990) thereby leading to a greater foliar glucose cost (e.g.Niinemets, 1997b).
Three species of Betula were used for the analyses.Betula nana L. is a very shade-intolerant shrub species (height 0´4± 1´2 m) with characteristic polycormic growth and with leaves 0´4±2 cm 2 in area on short, 0´5±1 mm long petioles, such that the leaves are seemingly directly attached to the stem.B. nana is widespread in tundra heaths (for further details, see Jonasson, 1981;de Groot et al., 1997), and as a relic of the last glaciation is restricted to raised bogs in Estonia (Dierssen, 1977).B. pubescens Ehrh.and B. pendula Roth.are closely related shade-intolerant tree species with leaves 5±20 cm 2 in area typically on 0´4±2 cm long petioles.Both species are pioneering trees, but B. pendula colonizes drier habitats than B. pubescens, which often dominates on wet soils (Atkinson, 1992).In Estonia, the maximum height of B. pendula is 35 m and that of B. pubescens 25 m (Masing, 1990).Because all the studied species coexisted in the most nutrient-limited wet habitat, but the competitive potential of B. nana was apparently lower at greater nutrient availabilities, we suggested that foliage characteristics adjust less plastically to improved nutrient availability in this than in the other species.The hypothesis that species responsiveness to nutrient availability depends on species growth potentials was posed by Coomes and Grubb (2000), but conclusive data to test the hypothesis are unavailable.Given that species success in nutrient-limited open environments is mainly driven by root characteristics, foliar variables of coexisting species may largely differ in these habitats.In contrast, in nutrient-rich systems, there is strong evidence of a functional convergence of foliage structural and physiological characteristics of coexisting species (Niinemets and Kull, 1994;Ackerly and Reich, 1999).
We used foliar nitrogen content per unit dry mass (N M ) as an estimate of internal plant nutrient availability.Plant nutrient contents depend not only on soil nutrient availability, but also on the ef®ciency and rate of nutrient uptake, and on nutrient partitioning within plant, as well as on the growth rates of the plant (Chapin et al., 1986;Hilbert, 1990;Oren and Sheriff, 1995).Thus, internal N concentrations control nutrient-related modi®cations in foliage structure and fractional biomass allocation, rather than soil-available N concentrations (Hirose, 1988;Levin et al., 1989;Hilbert, 1990;Tan and Hogan, 1998).Total tree height (H) was employed as a second independent variable to account for plant size related allocation and morphology differences.).The peat layers were extensiveÐup to 8 m in the centre of the bog, and in general more than 4 m at the margins (Veber, 1974).The sparse tree layer (200 trees/ha) with an average height of approx.1±2 m and an average age of 50±100 years was dominated by Pinus sylvestris L. and B. pubescens.Excavations of the entire root systems at the site demonstrated that the roots mostly occupied the upper peat horizons of 5±15 cm, and did not penetrate the dense peat layers.The average characteristics of the upper peat layers are reported in Table 1.Fertilizer experiments in temperate raised bogs have demonstrated that plant growth in these habitats is strongly curtailed by low soil nutrient availability.The data demonstrate that nitrogen and phosphorus are in especially short supply (Paavilainen, 1980;Damman, 1986;Fine Âr, 1992).
Betula pubescens was also investigated at two adjacent fell sites in the Viimsi peninsula.Both sites were previously dominated by Picea abies (L.) Karst.and were on acidic, sandy, podsolic soil with a raw humus horizon (Table 1), with the mineral soil layers starting at approx.20 cm depth.The ®rst fell site (Viimsi-1, 59°32¢N, 24°52¢E) was 2 years old, and ®rst and second year seedlings (24 trees, H = 0´134 T 0´023 m) were investigated at this site.The second clearing (Viimsi-2, 59°33¢N, 24°51¢E) was 5 years old, and 2±4-year-old saplings (20 trees, H = 0´67 T 0´07 m) were sampled.In both the bog and the fell sites, the light availability at the top of the sampled plants was always more than 90 % of that in a completely open location.We also included another set of trees of B. pendula (®ve trees, H = 1´10 T 0´21 m) sampled in Voore, Estonia (58°44¢N, 26°45¢E; Kull and Niinemets, 1993).The 30±35 m tall, and approx.100 year-old forest was dominated by P. abies, and had a vigorous deciduous shrub understory.The soil was a brown pseudopodzol with a mull type A horizon (Table 1).A thorough description of this study site is provided by Frey (1977).The light conditions above each sapling were assessed by hemispherical photography (Kull and Niinemets, 1993).The trees included in the current study grew in recent clearings or at the edges of the forest, and received at least 85 % of above-canopy irradiance.
Overall, the trees of different species that were sampled in various sites were of similar size.According to an ANOVA (analysis of variance), mean tree height was lower than in other cases for the ®rst and second year seedlings of B. pubescens sampled in Viimsi-1 (P < 0´001), and was not statistically different between other site±species pairs (P > 0´05).

Plant and foliar sampling: foliage morphological parameters
Plant sampling was conducted in July 1999 at Endla, September 1999 at Viimsi and August 1989 in Voore.There were no signs of foliar senescence during plant sampling.In Endla and Viimsi, the trees were excavated carefully to yield a reliable estimate of both below-and above-ground plant biomass.All leaves were harvested, and ten leaves from each plant were randomly taken for foliar area measurements.The remaining leaves, stem, coarse (diameter >1 mm) and ®ne roots were dried at 75 °C for at least 48 h, and weighed.The sample leaves were scanned by a Hewlett±Packard scanner at a resolution of 1200 d.p.i., converted to 1-bit bitmap, and the area of the sample leaves was measured by self-developed computer software.The sample leaves were also dried and weighed, and the leaf dry mass per unit area (M A ) was calculated.Total foliar area per plant was determined as the ratio of total leaf dry mass to M A .Leaf area ratio (LAR, cm 2 g ±1 ) was further computed as the ratio of total plant foliar area to total plant dry mass.
Given that the foliar characteristics of the plants excavated in Endla and Viimsi were averages for the entire plants, within-plant shading was likely to be considerable, and may have a critical affect on the correlations.To clarify the contribution of intra-plant shading to the relationships observed, we sampled the uppermost leaves of 35 separate plants in Endla.Because the top leaves were fully exposed, foliar characteristics could be studied without the interfering self-shading effects.Although we are not quantifying the degree of within-plant shading this way, the ®nding of the same correlations with the leaf characteristics of the uppermost leaves as with the whole-plant average leaf characteristics indicates that these correlations cannot be attributed to modi®ed within-plant shading alone.In Voore, only the topmost four to seven exposed leaves were sampled.For the top leaves, the circumference of each leaf was digitized, and the area was calculated as described by Niinemets et al. (1999).

Foliage and soil chemical analyses
Samples for soil analyses were taken from the A (A 0 ) horizon, where most of the roots of the sampled seedlings and saplings were found (5±10 cm for Endla, 5±15 cm for Viimsi and 3±20 cm for Voore).A sample for chemical analysis consisted of all leaves per plant (excavated plants in Endla and Viimsi) or of the uppermost completely exposed leaves in Endla (35 separate plants) and Voore (®ve separate plants).
For most samples, nitrogen and carbon contents were estimated by gas chromatography after combustion of the sample at >1000 °C in oxygen (elemental analyzer CHN-O-Rapid; Foss Heraeus GmbH, Hanau, Germany).Phosphorus contents were determined by inductively coupled plasma emission spectroscopy (Integra XMP; GBC Scienti®c Instruments, Melbourne, Australia) after digestion of pulverized leaves in 65 % HNO 3 .For some samples, standard Kjeldahl digestion was applied, and N content was determined by the indophenol method and P content by the molybdenum blue method (Grimshaw et al., 1989).Extensive comparisons of samples from different sites by various analytical routines indicated that the different methods gave similar results for both N and P with the differences `1 %.Ash content was estimated after the combustion of the sample in a muf¯e furnace at 500 °C for 3 h, and mineral content (A) calculated taking the fraction of minerals in ash as equal to 0´67 (Vertregt and Penning de Vries, 1987).
For the Endla and Viimsi sites, soil pH (1 g dry soil per 5 ml solution) was measured in distilled water (pH H 2 O ) and in 1 M KCl solution (pH KCl ).Soluble P [ammonium lactate± acetate-soluble phosphorus (AL-method, Swedish standard SS 02 83 10; Egner et al., 1960)] was determined by the molybdenum blue method after extraction with an aqueous solution (pH = 3´75) consisting of lactic (0´1 M) and acetic (0´3 M) acids, and ammonium acetate (0´1 M).The detailed soil characteristics and methods for chemical analyses of soils in the Voore site are reported by Ko Älli and Ka Èhrik (1970), Reintam (1970) and Kull and Niinemets (1993).

Leaf construction cost
Mineral-free carbon content (C A , g g ±1 ) was found as , where C m is total leaf carbon content per unit dry mass, and an estimate of leaf construction cost [G, g glucose (g dry mass) ±1 ] was computed as (Vertregt and Penning de Vries, 1987;Poorter, 1994;Niinemets, 1997b): (1) The primary assumption of eqn ( 1) is that the source of foliar nitrogen is ammonium or nitrogen in organic form, or that all nitrate is reduced in the leaf (Poorter, 1994;Niinemets, 1997b).Given that ammonium and amino acids are the predominate nitrogen source in bogs and raw humus soils (Chapin et al., 1986;Hobbie et al., 2000), we consider this assumption justi®ed.

Estimations of foliage photosynthetic electron transport rate
In Endla (Table 1), chlorophyll ¯uorescence measurements were conducted in the uppermost leaves with a pulseamplitude modulated ¯uorometer PAM-2000 (Heinz Walz GmbH, Effeltrich, Germany).Natural illumination was used for most measurements, but on cloudy days, the internal halogen lamp of the PAM-2000 was used as the source of actinic light.The leaf was held in a beam irradiance of 1200±1600 mmol m ±2 s ±1 , which was saturating for photosynthetic electron transport in the studied species.Pulses of white light of 8000 mmol m ±2 s ±1 were applied to determine the maximum ¯uorescence yield of light-adapted sample (F m ¢) as detailed by Niinemets et al. (1998b).From these measurements, the effective quantum yield of the light-adapted sample (F PSII ) was computed as (F m ¢ ± F)/ F m ¢, where F is the ¯uorescence yield in actinic light (Schreiber et al., 1994).Assuming that both Photosystems I and II absorb equal amounts of light, an estimate of the whole chain electron transport rate (J) was computed as described by Genty et al. (1989): ( where Q is the quantum ¯ux density and Q is the leaf absorptance.A constant value of 0´84 was used for Q in the current study.Average (T s.d.) leaf temperature was 27´7 T 3´1 °C during the ¯uorescence measurements and no attempt was made to standardize the observed values of J to a common temperature.

Statistical analysis
The maximum height of the excavated plants was only 1´7 m, but the height range of the studied plants was more than 50-fold, and signi®cant effects of tree height were found on both foliar morphological and plant allocation parameters.Because tree height and foliar nitrogen content were not strongly correlated (Fig. 1), the effects of total tree height and N could be studied independently.To improve the normality of the residuals, tree height, average leaf area, average dry mass and leaf area ratio were transformed to logarithms before the statistical analysis was performed.Simple and multiple linear regression analyses were employed to examine the dependencies of foliar and plant characteristics on total tree height and foliar nitrogen content.The species and site effects were investigated either by analyses of variance (Site and Species as main effects) when logH and N effects were non-signi®cant or by analyses of covariance (logH and N as covariates), and the means were separated by the Bonferroni test.Preliminary analysis indicated that the Site as well as Site Q N and Site Q logH interactions were generally non-signi®cant pendula, Table 1).Average foliar nutrient contents (Table 1) provided the primary criterion to distinguish the sites on the basis of nutrient availability (fertile vs. infertile).The same symbols are used for the Viimsi data because (1) the primary difference between the Viimsi sites (Viimsi-1, and Viimsi-2, Table 1) was only the age of the clear fells, and (2) the sites did not differ in measured soil characteristics and foliar nutrient contents.The linear correlation is non-signi®cant (r = 0´21, P > 0´06 for all data pooled).
(P > 0´05), indicating that Species, logH and N were the primary determinants of foliar characteristics.Given the low explaining power, the Site effects as well as the third-order interactions were suppressed in the ®nal ANOVA model.Because of differences in species abundance and dispersal patterns at the various sites, different numbers of plants were available for each species±site pair.In particular, the number of plants of B. pubescens was considerably larger than that of other species.This lack of balance may have potentially altered interspeci®c comparisons and the conclusions with respect to the signi®cance of the pooled regressions.To test for these possibilities, we constructed another dataset by randomly removing the data of B.
pubescens to give a total of 16 observations for this species.Although this procedure altered the statistical signi®cance of various differences and relationships, all multiple correlations were qualitatively the same for this and for the entire dataset, indicating that the pooled regressions were not affected by the lack of balance in the data.

Differences in the soil nutrient availability between the sites
The measured soil characteristics between the bog (Endla) and the fell sites (Viimsi-1 and Viimsi-2) were not statistically different.However, the upper soil horizons at the Voore site were signi®cantly less acidic, with lower C/N molar ratio compared with the other sites (Table 1).Nevertheless, foliar nitrogen and phosphorus contents (Table 1) suggested that the nutrient availability was signi®cantly larger at the fell sites than in the bog, possibly because of intensive soil mineralization following felling, or because the seedlings and saplings had access to deeper soil mineral horizons at these sites.
We observed a large within-site range in foliar nutrient contents, especially in the bog where foliar N contents ranged from 1´21 % to 2´74 %.Non-quantitative observations suggested that these differences were tightly related to the microheterogeneity in topography (e.g. the hummock± hollow gradient characteristic of the bogs, Titus and Wagner, 1984), and accordingly, in the water table level.In general, leaves with greater nutrient contents were sampled at higher locations, where the soil mineralization rates were likely to be higher.
According to Wassen et al. (1995), foliar P/N molar ratios >0´032 indicate primary N de®ciency, and P/N ratios <0´02 indicate P de®ciency in peat and raw humus soils.Given that the lowest P/N ratio observed in our study was 0´025 (Table 1), we conclude that the vegetation was primarily Nlimited in all sites.
Average leaf area and mass in relation to species, tree size and foliar nitrogen content Average leaf area was positively correlated with foliar nitrogen content per unit dry mass (N M ) in all species (Table 2, r 2 = 0´49, P < 0´001 for all data pooled).Taller trees also had larger leaves such that total tree height and S were positively related (Table 2, r 2 = 0´08, P < 0´002 for all data pooled).Given that logH and N M were essentially independent (Fig. 1), inclusion of both explaining variables into a multiple regression generally resulted in a higher fraction of explained variance than the separate regressions (Table 2, Fig. 2).
First ignoring the possible interactions, both the one-way analyses of variance (Table 3) and covariance (common slope model with logH and N M as covariates) suggested that at a common height and foliar nitrogen content, B. nana had the smallest leaves, B. pubescens intermediate, and B. pendula the largest leaves (P < 0´001; Table 3).Thus, interspeci®c variability was an important determinant of leaf size.However, further analysis of the species effects by a separate slope model indicated that the interaction term Species Q N was also signi®cant (P < 0´001, separate slope analysis of co-variance), as well as the interaction term Species Q logH (P < 0´05).These signi®cant interactions were indicative of lower responsiveness of leaf area in B. pubescens to N M , and logH (lower slope values, Table 2).
The species ranking was also identical in terms of average leaf dry mass (Table 3).As with the leaf area, the responsiveness of leaf dry mass to N M and logH was lower in B. pubescens than in the other species (P < 0´03 for Species Q N and P < 0´05 for Species Q logH interaction).

Effects of tree size and nitrogen content on leaf dry mass per unit area (M A )
Increases in the area of individual leaves with N M were larger than increases in the dry mass, such that M A scaled negatively with nitrogen content per unit dry mass (Fig. 3A, Table 2).Similarly, lower responsiveness of leaf area than leaf mass to logH led to a positive relationship between M A and logH (Fig. 4A, Table 2).The scaling of M A with N M was TAB L E 2. Species-speci®c regressions between foliar structural characteristics, and foliar nitrogen content and total tree height Nitrogen content (N M , %) Tree * logS = logarithm of average leaf size (cm 2 ), logM = logarithm of average leaf dry mass (g), M A = leaf dry mass per area (g m ±2 ), log(LAR) = logarithm of leaf area ratio (cm 2 g ±1 ), F L = fraction of foliar biomass in leaves (g g ±1 ), ² Both nitrogen content and logH included as explaining variables.³ LogH was not statistically signi®cant in the multiple regression (P > 0´4).§ N M was not statistically signi®cant in the multiple regression (P > 0´9).
signi®cant only in B. pubescens, and scaling with logH in B. nana and B. pubescens (Table 2).Nevertheless, the species effect (P > 0´6) as well as Species Q N (P > 0´4) and Species Q logH (P > 0´8) interactions were non-signi®cant, indicating that despite the data scatter in B. nana and B. pendula, they adhered the same overall trend (Fig. 4A).
All plants included in the study were exposed to more than 85 % of the light in a completely open location.However, the values of M A depicted in Figs 3A and 4A were obtained either as an average of all leaves per plant or as the average of uppermost exposed leaves.To test the possibility that the negative scaling of M A with N M resulted from greater self-shading because of a larger foliar area in plants with better N nutrition, we also computed the regressions separately for the whole plant average M A and for the uppermost M A values.In both sets of data, M A was negatively correlated with N M and positively with logH.For all species pooled, r 2 = 0´55 (P < 0´001 for both independent variables) for the uppermost leaves, and r 2 = 0´61 (P < 0´001) for the whole plant averages.Thus, increased self-shading within the canopy did not provide the only explanation for the strong negative relationships between M A and N M (Table 2, Fig. 4A).
Leaf nitrogen content per unit area (N A = N M ´MA ) and M A were positively related (Fig. 3B).However, because of the differences in N M and a negative relationship between M A and N M (Table 2, Figs 3A and 4A), the intercept of the M A vs. N A relationship was larger in the leaves with lower N M (P < 0´001 for the site effect, Fig. 3B).Despite N M being independent of tree height (Fig. 1), positive effects of logH on M A (Fig. 4A) also led to a positive correlation between N A and logH (r 2 = 0´43, P < 0´001).

Dependence of leaf area ratio (LAR) and fraction of plant biomass in leaves (F L ) on tree size and nitrogen content
Leaf area ratio (total foliar area per total plant dry mass) was negatively related to logH and positively to N M in B. pubescens, but not in B. nana (Table 2, Fig. 4B).However, covariance analysis demonstrated that the values of LAR of B. nana overlapped with those observed in B. pubescens.Both interaction terms, Species Q logH and Species Q N, were non-signi®cant (P > 0´6 for both) according to the separate slope ANCOVA model, and the main effect, Species, was insigni®cant according to the common slope model (P > 0´08).Thus, despite the very different plant architecture, the LAR values of both species were similar.
The fraction of plant biomass in foliage (F L ) was also negatively related to logH and positively to N M (Table 2, r 2 = 0´38, P < 0´001 for all data pooled).Again, the interaction terms were non-signi®cant in the separate slope model (P > 0´7 for both) and the species effect was non-signi®cant according to the following common slope analysis (P > 0´1).
Given that LAR is dependent on both M A and F L (LAR = F L /M A ; Fig. 5), the plasticity in both of its components with respect to N M and logH controls the variation of LAR along gradients of nutrient availability and with tree size.To determine the extent to which the N M -and logH-related variability in LAR can be ascribed to variation in F L and M A , we calculated the expected values of F L and M A using the minimum and maximum measured values of logH and N M in interspeci®c regressions (Fig. 4A).For the observed range of logH (±1´50 . . .0´23), the expected value of F L varied 6´7-fold, and that of M A only 1´9-fold.For the observed range in N M (1´21 . . .3´36), the expected value of F L varied 1´9-fold, and that of M A 1´3-fold.Thus, we conclude that F L responded more plastically than foliar structure to both logH and N M , and was, accordingly, a more important determinant of LAR.

Foliar photosynthetic characteristics and construction cost vs. M A , nitrogen and tree height
The light-saturated rate of photosynthetic electron transport per unit dry mass (J M ) was positively related to foliar nitrogen content (Fig. 6A), whereas the regressions were signi®cant for each individual species (P < 0´05), as well as for all data pooled.However, J M scaled negatively with M A (Fig. 6B).The rate of photosynthetic electron transport per unit area (J A ) correlated also positively with N A (Fig. 6C), but because of the negative correlation between J M and M A , J A (J A = J M ´MA ) was independent of M A (Fig. 6D).A multiple regression analysis of J M vs. M A and N M (r 2 = 0´73) indicated that the negative scaling of J M did not solely result from N M effects on M A (Fig. 3A), but both M A (P < 0´001) TAB L E 3. Averages (T s.e.) of foliage structural and chemical characteristics and foliar construction cost: results of oneway analyses of variance* M A = leaf dry mass per unit area, LAR = leaf area ratio (total leaf area per unit total plant mass), G = leaf construction cost (eqn.1), n.d.= not determined.
* Each individual species-site pair was used as a factor level.Means followed by the same letter are not signi®cantly different at P < 0´05.
² The variable was transformed to a logarithm before statistical comparisons.
and N M (P < 0´001) were signi®cant determinants of J M .The ratio of J M to N M was negatively related to M A (r 2 = 0´30, P < 0´001), further showing that the nitrogen use ef®ciency in photosynthesis declines with increasing M A .
Multiple linear regression of J M vs. N M and logH indicated that, at common N M , J M was lower in larger trees (r 2 = 0´68, P < 0´01 for logH), possibly because of a larger M A (cf. Figs 4A and 6B).However, J A did not depend signi®cantly on logH (P > 0´1) in the multiple regression of J A vs. N A and logH.None of the correlations highlighted differed between species (P > 0´5 according to commonslope ANCOVA analyses).
Foliar construction cost (G) tended to be higher in sites with greater nutrient availability (Table 3) and there was a positive relationship between G and N M for all data pooled (r 2 = 0´63, P < 0´001), but not with logH (P > 0´4).

Changes in leaf size with nitrogen status
Strong positive relationships between average leaf area and foliar nitrogen content were found in all species (Fig. 2, Table 2).According to laboratory studies, enhanced nitrogen availability increases both the rate of leaf expansion as well as cell division (MacAdam et al., 1989;Taylor et al., 1993;Roggatz et al., 1999) and leads to larger individual leaves (Karlsson and Nordell, 1989;Tra Âpani and Hall, 1996;Sims et al., 1998a).Whole plant experiments further indicate that, in addition to larger leaves, the number of leaves may also increase with increasing N availability (Ackerly and Bazzaz, 1995;Grubb et al., 1996).Although the same total foliar area may be formed by a fewer larger leaves or by a larger number of smaller leaves, we suggest that the plant strategy to increase the area of individual leaves with increasing nitrogen availability has distinctive ecological advantages over the strategy to increase leaf number only.First, height growth becomes an increasingly more important competitive attribute as the nutrient limitations decrease, because enhanced nutrient supply leads to augmented light competition (Tilman, 1986;Gleeson and Tilman, 1992).As the interspeci®c correlations suggest, the branching requirements are less for larger leaves (Givnish, 1984;King, 1991King, , 1998)).Thus, larger leaves allow a greater fractional investment of plant biomass in height growth relative to the lateral growth (King, 1991).In addition, simulation studies demonstrate that with increasing length of leaf blades and petioles, the degree of foliar self-shading by other leaves and stem decreases, improving the light interception ef®ciency of unit foliar area (Takenaka, 1994).All the species of Betula studied here possess short petioles, and increases in leaf lamina size should signi®cantly improve the light interception ef®ciency in these species.
Given this ecological reasoning, interspeci®c differences in leaf size (Table 3) may partly explain the contrasting competitive ability of the studied species (B.nana < B. pubescens < B. pendula) in nutrient-rich early-successional habitats.Especially in B. nana, the extremely small leaf size and lack of petioles leads to a large fraction of foliar area that is shaded by the stem.We suggest that this, in combination with genetically constrained height growth potential, is the primary factor limiting the competitive potential of this relic of the last glaciation period in nutrientrich habitats.

Leaf dry mass per unit area in relation to foliar nitrogen
In agreement with the contrasting patterns of nitrogen content vs. M A relationships of previous studies, M A vs. N M relations were weaker than leaf size vs.N M dependencies (Table 2).Nevertheless, there was a statistically robust negative correlation between N M and M A in B. pubescens (Table 2).Although the M A of both B. nana and B. pendula was non-signi®cantly related to N M , these data ®t the same regression surface with B. pubescens (Fig. 4A), indicating a functional convergence in foliar characteristics.These non-signi®cant relationships may be a manifestation of lower plasticity, but may have also resulted from clustering of M A data of B. nana and B. pendula at relatively low N contents (Fig. 4A).
The relationship between N M and M A was qualitatively identical when a whole-plant average M A or M A of the uppermost leaves was used in the regressions.This consistency demonstrates that the negative effect of site nutrient conditions on M A was not solely caused by increased self-shading within the canopy.Given that self-shading was ruled out as the only determinant of low M A at higher N M , it is pertinent to ask by which additional mechanisms the changes in M A in B. pubescens may have occurred along the nutrient gradients.Fundamental insight into the nutrient availability effects on M A may be obtained by distinguishing between the two components of M A , leaf thickness (T) and density (D, M A = T´D), that may vary independently (Witkowski and Lamont, 1991;Niinemets, 1999).The strong positive relationship between lateral expansion growth and N M suggests that the growth in thickness may also respond to enhanced N availability in a similar manner.Such a positive relationship has been found in some investigations (Jurik et al., 1982;Thompson et al., 1992;Sims et al., 1998a), and may be the consequence of larger cells in leaves produced at higher N availability (Roggatz et al., 1999).Yet, there are consistent negative effects of N on D (Witkowski and Lamont, 1991;Niinemets et al., 2001, recalculated from Jurik et al., 1982;Thompson et al., 1992) that are generally larger than the positive effects of N on T, such that the product variable, M A , decreases with increasing nutrient availability (Jurik et al., 1982;Thompson et al., 1992).Data demonstrate that the number of cells per unit area is larger (Roggatz et al., 1999), and the fraction of intercellular air space lower (Jurik et al., 1982) in nitrogen-limited leaves, providing an explanation of greater density of these leaves.Moreover, starch (Waring et al., 1985;Bowler and Press, 1996;Sims et al., 1998b) and lignin (Waring et al., 1985) contents may accumulate, further increasing leaf density at a common thickness.

Modi®cation of leaf size and M A by tree size
Average leaf area was positively related to total tree height in three Betula species (Fig. 2), as in previous studies on young trees of Betula pubescens ssp.tortuosa (Sveinbjo Èrnsson, 1987;Senn et al., 1992).Apart from the ecological bene®ts of larger leaves in terms of the ef®ciency of height growth, shorter plants have branches and stems with lower diameter.Accordingly, both mechanical and hydraulic constraints may be responsible for the limited maximum leaf size in smaller plants.Leaf size generally scales positively with the diameter of supporting branchlet (White, 1983).Although we observed a continuous increase in leaf size with increasing plant height (Fig. 2), other investigations conducted over a larger range of plant height and age indicate that the leaf size vs. plant age or height relationship may be a curve with a maximum, with the leaf size either levelling off or decreasing after the leaves have reached the maximum size (Montfort and Mu Èller, 1951;Steele et al., 1989).Increasing hydraulic limitations and F I G . 6. Relationships between the rate of light-saturated photosynthetic electron transport (J) per unit leaf dry mass in dependence on leaf nitrogen content per unit dry mass (A) and M A (B); and J per unit leaf area in relation to nitrogen content per unit area (C) and M A (D).The rate of photosynthetic electron transport was determined from chlorophyll ¯uorescence measurements (eqn.2).The data were ®tted by linear regressions.
Symbols as in Fig. 1.
greater water stress in taller trees (Ryan and Yoder, 1997;Mencuccini and Magnani, 2000) may provide an explanation for this curvilinearity.
In contrast to the non-linear relationships between leaf size and plant height, M A increases linearly with increasing tree size (Steele et al., 1989;Niinemets 1997a;Niinemets et al., 1999), as was also observed in our study in B. nana and B. pubescens (Table 2, Fig. 4A).Similarly to M A increases in nutrient-limited leaves, the positive scaling of M A with tree size primarily results from increases in leaf density (Niinemets, 1997a(Niinemets, , 1999) ) rather than from increases in thickness (but cf.Malkina, 1983;Niinemets et al., 1999).Because increases in density are compatible with larger tissue elastic moduli that allow a greater change in leaf water potential with a lower tissue water loss (Niinemets, 2001), it has been suggested that the tree-size related increase in M A is an acclimation response to a more severe water stress in the canopies of larger trees (Niinemets 1997a).

Biomass partitioning in relation to N and tree size
Increases in M A with increasing tree size and with decreasing nitrogen content (Fig. 4A) in B. pubescens evinced that construction of extensive foliar area in terms of foliar biomass investment in leaves may be more expensive in taller trees and at low N (Figs 4B and 5).Our study indicates that, in all species, changes in LAR (LAR = F L / M A ) resulted primarily from modi®cations in fractional biomass allocation to leaves, F L , rather than in M A .Because the plants grow faster at a greater nutrient supply, the plants of the same age are also larger at higher nutrient availability.Given the scaling of plant size with nutrient availability and M A with tree size (Fig. 4A), we hypothesize that the plastic modi®cations in M A in response to nutrient availability may be constrained by tree size effects on M A .
The evidence that N M effects on M A , LAR and F L were non-signi®cant in B. nana but signi®cant in B. pubescens (Table 2) con®rms quantitatively the suggestion that more competitive species with larger growth rates are more responsive to nutrients (Coomes and Grubb, 2000).However, the non-signi®cant effect of N M on M A in B. pendula apparently contradicts this explanation.Nevertheless, although we did not study LAR and F L in B. pendula, previous investigations have demonstrated that LAR and F L are plastically modi®ed along nutrient availability treatments in this species (McDonald et al., 1986(McDonald et al., , 1992)).Thus, the allocation patterns suggest that B. pendula ®ts the general hypothesis of a larger plasticity in more competitive species.
Nitrogen status and tree size vs. the foliage photosynthetic characteristics and the construction cost As observed previously (for a review, see Field and Mooney, 1986), increased foliar N contents strongly enhanced foliar photosynthetic capacities (Fig. 6A, C).Studies dealing with the leaf structure vs. photosynthetic capacity relationships along light gradients have further found a strong positive correlation between M A and foliage photosynthetic capacity per unit area (e.g.Ellsworth and Reich, 1993;Niinemets et al., 1998a).This correlation has been interpreted as indicative of greater thickness of leaves with larger M A (Niinemets et al., 1999), and consequently, characterizing an accumulation of photosynthetic biomass per unit leaf area with increasing M A (Niinemets et al., 1998a).However, we found no relationships between the capacity of photosynthetic electron transport (J) per unit area (J A ) and M A (Fig. 6D), and a negative correlation between J per unit dry mass (J M ) and M A (Fig. 6B).We hypothesize that the negative scaling of J M with M A at a common foliar nitrogen content resulted from increases in leaf density with decreasing nutrient availability.Leaves with a greater density have larger intercellular gas-and liquid-phase transfer resistances to CO 2 (Syvertsen et al., 1995), and consistently lower photosynthesis rates at common stomatal limitations (Niinemets, 1999).
Studies have brought evidence of a negative scaling of foliar nitrogen contents per unit dry mass (Niinemets, 1997a) and leaf assimilation potentials with increasing tree size (Grulke and Miller, 1994).Although the positive effects of tree size on M A led to a greater nitrogen content per unit leaf area in our study, J A was independent of tree size, and J M decreased with increasing tree size.Greater density and diffusive limitations in leaves with greater M A may also be responsible for the negative effects of tree height on J M in the three Betula species studied.
Despite the extensive modi®cations in M A (Fig. 4A), foliage construction cost (G) changed little across different foliar morphologies.Although carbon-rich lignin (63´3 %, calculated according to Nimz, 1974) and tannins may accumulate at lower nutrient availabilities (Waring et al., 1985), there is also evidence of higher contents of nonstructural carbohydrates in nutrient-limited leaves (Waring et al., 1985).Given that the non-structural carbohydrates have low carbon content (44´4 % for starch), simultaneous accumulation of both the expensive and cheap compounds at low nutrient availability may be a reason for lower G at low nutrient availability.Yet, we found that G was strongly related to N M , which provides an estimate of foliar protein contents.Because proteins are rich in carbon (53´5 %), the construction of leaf photosynthetic apparatus with high N content is also costly in terms of carbon investments in unit foliar biomass.Although G should not always be related to site fertility (Chapin, 1989), our ®nding agrees with the majority of studies (La®tte and Loomis, 1988;Grif®n et al., 1993;Grif®n et al., 1996;Poorter and Villar, 1997) demonstrating a larger construction cost in leaves with higher N content.

C O N C L U S IO N S
The information summarized indicates that the effects of internal plant N on M A are complex in temperate deciduous woody species, and may partly be mediated through enhanced self-shading and increased plant size at higher nutrient availability.Yet, there are also important independent effects of nitrogen on foliar structure, which seem mainly to result in denser leaves at lower N. Increases in density may provide an explanation for a stronger reduction in foliar photosynthetic capacities with decreasing nitrogen content than may be predicted from decreases in N M alone.High-density mesophyll is generally an adaptation to waterlimited environments such as Mediterranean communities and semi-deserts (Niinemets, 2001), and the functional signi®cance of dense leaves in nutrient-limited habitats remains to be discovered.
The biomass allocation responds more readily to differences in N than M A , indicating a larger plasticity in allocation than in foliar structure.However, as the trees increase in size, the allocation in foliage consistently decreases.Thus, the positive effect of high internal nutrient availability on foliar biomass investment in leaves becomes buffered by concomitant increases in plant size that reduce F L .Given the strong controls of plant size, the studies on N in¯uences on foliar structure and plant biomass allocation should also include plant size as a covariate in statistical comparisons.
M A T ER IA L S A N D M ET H O D S Study sites Betula nana L. (16 plants with average T s.e. total height, H, of 0´39 T 0´05 m), B. pubescens Ehrh.(43 trees, H = 0´70 T 0´08 m) and B. pendula Roth.(six trees, H = 0´58 T 0´15 m) were studied at a nutrient-poor site on Sphagnum peat in Ma Ènnikja Èrve raised bog, Endla State Nature Reserve, Estonia (58°52¢N, 26°15¢E

F
I G . 2. Dependence of leaf size on foliar nitrogen content and on total tree height in Betula nana, B. pendula and B. pubescens.All data were ®tted by a single multiple linear regression: logS = ±0´190 + 0´358logH + 0´447N M (r 2 = 0´60, P < 0´001 for both logH and N M , P > 0´1 for the intercept).Symbols as in Fig. 1.F I G . 1. Scatter plot of foliar nitrogen content per unit dry mass vs. total tree height.The data were sampled from the bog (poor site) and from the clear fell sites in Viimsi (B.pubescens) and forest edges in Voore (B.

F
I G . 3. Correlations between leaf dry mass per unit area (M A ) and leaf nitrogen content per unit dry mass (A) and N per unit area (B).In A, all data were ®tted by a single linear regression (r 2 = 0´33, P < 0´001).In B, separate regressions were ®tted for all sample points from the infertile site (upper regression line, r 2 = 0´31, P<0´001) and for B. pubescens from the fertile site (lower regression line, r 2 = 0´85, P<0´001).Symbols as in Fig.1.FI G . 4. Effects of leaf nitrogen content and total tree height on M A (A) and on leaf area ratio (B).The regression surfaces were ®tted through all data: M A = 123´4 + 23´2logH ± 18´8N M (r 2 = 0´64) for A, and log(LAR) = 0´622 ± 0´340LogH + 0´277N M for B (r 2 = 0´49).All regression coef®cients were signi®cant at P < 0´001.Symbols as in Fig.1.

F
I G . 5. Leaf area ratio in relation to leaf dry mass per unit area.The regression was ®tted through all data.Symbols as in Fig.1.
TAB L E 1. Site average (T s.d.)* chemical characteristics of the upper soil horizon and average foliar chemical data