Introduction

A primary cost for obtaining CO2 for photosynthesis is the evaporation of water from leaf surfaces. Therefore, plants must continuously optimise their stomatal openness to enable the acquisition of CO2 on one hand, and to prevent excessive water loss from foliage on the other hand. However, leaf water status depends not only on leaf diffusive conductance and atmospheric evaporative demand but also on leaf water supply. Thus, the ability of plants to supply water to foliage is ultimately associated with their productivity, competitiveness and survival. A number of studies (e.g., Köstner et al. 2002; Macinnis-Ng et al. 2004; Santiago et al. 2004; Renninger et al. 2006) have confirmed a relationship between plant hydraulic properties and stomatal conductance, photosynthetic capacity and growth rate. Differences in the competitive capability of conifer and angiosperm species are associated with their respective xylem physiology (Brodribb et al. 2005; Pittermann et al. 2005).

Leaf water supply depends primarily on soil moisture availability, hydraulic conductivity of the vascular tissues, sapwood area to leaf area ratio (Huber value, HV), and driving forces (i.e., water potential gradient) between the soil and leaves. Hydraulic conductivity is not uniform across the whole vascular transport path; we focus on branches that contain a major portion of the stem resistance (Yang and Tyree 1994). Maximum xylem hydraulic conductivity is primarily determined by wood anatomy, particularly on the distribution of conduit inner diameters (Domec et al. 2006; Hacke et al. 2006; Renninger et al. 2006): according to the Hagen–Poiseuille law, conduit hydraulic conductivity is proportional to its diameter raised to the fourth power.

The hydraulic limitation hypothesis (HLH) proposes that the decline of forest productivity with age is a consequence of the loss of whole-plant and leaf-specific hydraulic conductance with tree height resulting from a rise in friction and gravitational potential, and increased investment in non-photosynthetic tissues (Ryan and Yoder 1997; Woodward 2004; Mencuccini et al. 2005). As trees grow taller, their stomatal conductance and photosynthesis wane, and, as a result, the height growth declines. A review of 51 studies confirmed that hydraulic limitation of gas exchange with increasing tree size is a common—but not completely universal—phenomenon (Ryan et al. 2006).

The HLH was proposed to provide a mechanism for gas exchange and growth limitation observable at the whole-tree level. Opposite patterns of hydraulic capacity have often been revealed within crowns of both coniferous and broad-leaved trees: specific hydraulic conductivity of branches (Protz et al. 2000; Lemoine et al. 2002; Jerez et al. 2004; Burgess et al. 2006) or shoot hydraulic conductance increased with branch height in the crown (Aasamaa et al. 2004). Also apparent soil-to-leaf hydraulic conductance (G T), defined as a current transpiration rate divided by water potential drop from soil to leaves (Meinzer et al. 1995; Wullschleger et al. 1998), has been shown to increase from basal to top branches (Sellin and Kupper 2004, 2005a, 2007a). Thus, qualitatively different phenomena can be observed within a crown. Branch autonomy may promote environmental screening and enhance differential growth by optimising the use of available light resources (Henriksson 2001). Extensive studies on silver birch (Betula pendula Roth) suggest that water flow from soil to shade foliage located at the crown base needs to overcome greater resistance than flow to sun foliage at the tree top (Sellin and Kupper 2005a, b). The mean G T in birch was 1.7 times higher for the upper canopy compared to the lower canopy. In a recent paper, Sellin and Kupper (2006) revealed a pronounced vertical trend in Huber values: at the whole-branch level, HV increased basipetally within the crown. Thus, sapwood to leaf area ratios of the branches cannot account for the differences in G T at different crown positions in silver birch. On the contrary, the higher HV observed in the lower-crown branches may be considered a means to compensate for differences in xylem conducting capacity between the lower and upper canopy. Hydraulic compensation by adjusting sapwood to leaf area ratio with increasing tree height has been well documented in multiple species (Ryan et al. 2006).

In this paper, we report the results of anatomical studies of branchwood in silver birch and discuss them extensively with respect to data published for other woody species. The main objective of this study was to gain a more detailed understanding of how different parts of crown compare anatomically and hydraulically, and their contribution to the distribution of transpirational fluxes within the canopy. Spatial patterns of anatomical and hydraulic characteristics of xylem within a crown were investigated from three perspectives: vertical variation, variation with distance from trunk, and variation depending on branch order. This paper addresses two specific objectives: first, to establish the variation in vessel diameter and density, the characteristics relevant to both hydraulic and mechanical properties of xylem (Preston et al. 2006), with respect to canopy position, and second, to test the hypothesis that vertical differences in G T within the crown result from a variation in specific conductivity of xylem. Distribution of hydraulic conductance within the crown is an essential factor driving canopy development during forest maturation. Protz et al. (2000) suggested that reduction in branch sapwood hydraulic conductivity over time limits survival of lower branches, and therefore controls the vertical crown recession in closed-canopy trees. On the other hand, forest management operations may have substantial consequences on plant hydraulics, affecting both growth rate of trees and wood characteristics (Renninger et al. 2006). In addition, this topic concerns the perspectives of our forests; in light of global climate change, hydraulic architecture of trees has considerable implications for forest condition and productivity over longer periods (DeLucia et al. 2000; Maherali and DeLucia 2000).

Materials and methods

Study area

The field studies were carried out in Järvselja Experimental Forest (58°16′N, 27°20′E, elevation 38–40 m) in eastern Estonia. The average annual precipitation is 650 mm; the average temperature is 17.0°C in July and −6.7°C in January. The growth season usually extends 175–180 days, from mid-April to October. The annual global short-wave radiation in the region averages 3,518 MJ m−2, and the annual radiation budget is 2,552 MJ m−2 (Russak 1990). Detailed climatic data on the study area were presented by Sellin and Kupper (2005b). The soil is a gleyed pseudopodsol formed on a loamy till with a large water storage capacity (Niinemets and Kull 2001).

The studies were performed in a mixed stand dominated by silver birch (B. pendula) and growing in a Vaccinium myrtillus site type. The total basal area is 33.2 m2 ha−1; the stand composition is: B. pendula 46%, Picea abies (L.) Karst. 44%, Pinus sylvestris L. 8%, and Populus tremula L. 2% of the total basal area. The mean (±S.E.) height of the birch trees is 18.1 ± 0.26 m, DBH 10.8 ± 0.21 cm, and live crown ratio is 0.40 ± 0.008.

Sampling of trees

We sampled five birch trees, 22.0–26.0 m in height, from which three branches were sawn off—one branch from the basal, middle and upper third of the crown. The diameter at the base and length of each branch were measured in the laboratory. At each primary branch, we selected one II order and one III order branch inserted in the former. The branch was then dissected and sampled at five points avoiding branch junctions:

Pr:

the primary branch at 1/10 of its total length;

Mi (or I):

the primary branch at ½ of its total length;

Di:

the primary branch at ¾ of its total length;

II:

the second-order branch in the middle;

III:

the third-order branch in the middle.

Exact sample locations were determined with a tape measure. From each sampling location a branch segment 3–4 cm in length was cut, the conducting xylem of which was identified by perfusing an aqueous solution of methylene blue (0.5% w/v) through the specimens by means of a vacuum pump. The sapwood cross-sectional area, foliage area situated acropetally in relation to the wood specimens, and Huber value (sapwood area to leaf area ratio, HV) were determined as described in Sellin and Kupper (2006). The wood specimens were dried at 80°C for at least 48 h and preserved for anatomical studies.

Wood anatomy and xylem conducting efficiency

The dried wood samples were boiled in distilled water for about 2 h to soften the wood and thin slices were cut with a razor blade. Cross-sectional surfaces of the slices were coated with gold in a sputter coating system and examined with a BS-301 scanning electron microscope (Tesla, Czech Republic). Three digital images (one at 300× and two at 500× magnification) covering one to seven outermost growth rings were taken of each wood specimen and the images analysed with a Scion Image (release Beta 4.0.2; Scion Corporation, USA) software. From the 500× images we measured minor and major vessel inner diameters (μm), estimated as lengths of the minor and major axes of the best-fitting ellipse, cross-sectional area of vessel lumina (μm2), and vessel density (mm−2). For most images the number of measured vessels ranged 120–170. Width of the growth rings (mm) was determined at the 300× magnification. Mean vessel diameter (D) was calculated as follows:

$$ D = \frac{{{\sum\nolimits_{i = 1}^n {{\sqrt {a_{i} b_{i} } }} }}} {n}, $$
(1)

where a i and b i are minor and major perpendicular lumen diameters of the ith vessel, and n is the number of the measured vessels. Vessel diameters were analysed for their frequency in 5 μm diameter classes (10.0–14.9, 15.0–19.9 μm, etc.) and for their relative contribution of each diameter class to the sum of all the conduits raised to the fourth power. On the basis of the vessel contribution to hydraulic conductance, we calculated a hydraulically weighted mean diameter (D h) for each sample (Sperry et al. 1994; Kolb and Sperry 1999):

$$ D_{{\text{h}}} = \frac{{{\sum\nolimits_{i = 1}^n {({\sqrt {a_{i} b_{i} } })^{5} } }}} {{{\sum\nolimits_{i = 1}^n {({\sqrt {a_{i} b_{i} } })^{4} } }}}. $$
(2)

To characterise the xylem conducting efficiency expressed per sapwood transverse area, a theoretical specific conductivity of the xylem (k t, kg m−1 s−1 MPa−1; sensu Tyree and Ewers 1991) was used. In fact, k t indicates theoretical lumen conductance, omitting the resistance between vessels. It was computed according to Lewis and Boose (1995):

$$ k_{t} = \frac{{\pi \rho }} {{64\eta A_{{\text{x}}} }}{\sum\limits_{i = 1}^n {\frac{{a^{3}_{i} b^{3}_{i} }} {{a^{2}_{i} + b^{2}_{i} }}} }, $$
(3)

where ρ is the density of water (kg m−3), A x is the sampled cross-sectional area of xylem (m2), and η is the dynamic viscosity of water (MPa s). Theoretical leaf-specific conductivity of the xylem (LSCt; kg m−1 s−1 MPa−1) was calculated as the product of k t and HV. The theoretical hydraulic characteristics were calculated for a reference temperature of 20°C.

Data analysis

Data analysis was carried out using Statistica, Version 6.0 (StatSoft, USA). In the analysis with respect to branch orders, we compared samples from middle part of the branches, thus the primary branches were represented by sample location Mi (i.e., Mi = I). To analyse the effects of branch position in the crown (lower, middle and upper crown), sample radial location along main branches (sample locations Pr, Mi and Di), branch order (sample locations I, II and III), as well as other branch variables on anatomical and hydraulic characteristics, an analysis of covariance (ANCOVA) was performed using the general linear models module. Branch position, branch order, and sample location were treated as fixed factors, while branch length, branch diameter measured under the bark (i.e., diameter of the xylem cylinder) at sampling points and mean width of the growth rings were treated as covariates. The covariates were included in the models to reveal possible effects resulting from variation in branch size and growth rate, which may affect the relationships of the categorical independents to the dependent variables. During the analysis, the models were modified by a step-by-step removal of insignificant covariates. Type IV sums of squares were used in the calculations. Normality and homogeneity of variances were checked using the Kolmogorov–Smirnov D-statistic and the Levene test, respectively. When appropriate, logarithmic (LSCt) or square root transformations (k t, vessel density) were applied to the data. Post hoc mean comparisons were conducted using the Tukey HSD test. Shape and strength of the bivariate relationships between the studied characteristics and independent variables were analysed by applying linear or non-linear regression procedures based on the least-squares method.

Results

Vertical variation

The lower, middle and upper thirds of the crown did not vary in vessel lumen diameters (D) in branchwood; D for the three canopy layers ranged from 22.6 to 23.3 μm. Diameter of individual vessels varied between 10 and 60 μm, the modal diameter class (23.6–25.3% of all vessels) for all three canopy positions was 15–20 μm. ANCOVA revealed a significant (Table 1, P = 0.025) effect of crown position on hydraulically weighted mean diameter (D h) of vessels. However, the Tukey test revealed no clear vertical trend within the canopy (Fig. 1a). Vessel density (VD) ranged from 122 to 702 mm−2 and vessel lumina covered 8.4–26.8% of the sapwood cross-sectional area in the whole data set. Neither VD nor relative area of vessel lumina (VA) depended on branch height within a crown as far as covariates were included in the ANCOVA model (Table 1). After the covariates were discarded, branch position turned out to be a significant (P < 0.05) factor: VD increased acropetally within the crown (Fig. 1b). Judging by the mean squares, mean width of the growth rings (W) accounted for 22.6% of the variation in VD. In the analysis of covariance, the influence of branch position had been hidden by effects of W, because width of the branch growth rings varied with canopy layers (P < 0.001), and W and VD were positively correlated (Pearson’s correlation coefficient r = 0.195, P < 0.05). Higher branch radial growth rate was also associated with larger proportion of vessels to the total cross-sectional area of xylem (Pearson r = 0.479, P < 0.001).

Table 1 Results of ANCOVA for the effects of branch position and other factors on anatomical and hydraulic characteristics of the branchwood (analyses across all sampling locations and branch orders)
Full size table
Fig. 1
figure 1

Spatial variation of mean hydraulic diameter of vessels (a), vessel density (b) and relative area of vessel lumina (c) within a crown: vertical variation (Lc, M c, U c—lower, middle and upper crown, respectively); radial variation (Pr, Mi, Di—branch proximal, middle and distal parts, respectively; variation depending on branch order (I, II, III—first-, second- and third-order branches, respectively). The bars indicate ±SE of the means; different letters at the columns denote statistically significant (P < 0.05) differences

Full size image

Specific conductivity of xylem (k t) did not vary with branch position (Table 1, Fig. 2a), whereas leaf-specific conductivity (LSCt) exhibited a decreasing trend with increasing branch height (Fig. 2b). Variation in k t can be divided into two components—mean hydraulic diameter and vessel density. To estimate combined effects of D h and VD on k t, we applied non-linear regression analysis and developed an empirical model. The best fit resulted from the following multiplicative function (Fig. 3):

$$ k_{t} = - 12.98 + D_{{\text{h}}} ^{{0.580}} {\text{VD}}^{{0.143}} ,\quad n = 150,\enspace R^{2} = 0.854,\enspace {\text{S}}{\text{.E}}{\text{.E}}{\text{.}} = 0.540, $$
(4)

where all three parameters were statistically highly significant (P < 0.001).

Fig. 2
figure 2

Spatial variation of theoretical values of specific conductivity (a) and leaf-specific conductivity of xylem (b) within a crown: vertical variation (L c, M c, U c—lower, middle and upper crown, respectively); radial variation (Pr, Mi, Di—branch proximal, middle and distal parts, respectively; variation depending on branch order (I, II, III—first-, second- and third-order branches, respectively). The bars indicate ±SE of the means; different letters at the columns denote statistically significant (P < 0.05) differences

Full size image
Fig. 3
figure 3

Theoretical specific conductivity of xylem (k t) as a function of mean hydraulic diameter of vessels (D h) and vessel density (VD)

Full size image

Specific conductivity of xylem also depended significantly (Table 1) on mean width of the growth rings; higher radial growth rate of branches was associated with higher hydraulic conductivity of xylem (Pearson r = 0.358, P < 0.001). The data analysis revealed a significant effect (P < 0.001) of the branch diameter (d b), measured at sampling points, on VD, VA as well as k t. D h, k t and LSCt increased, and VD declined with increasing d b (Fig. 4). Thus, D h and VD exhibited opposite trends with branch diameter, resulting from a strong negative relationship between vessel size and vessel density (Fig. 5).

Fig. 4
figure 4

Branch diameter measured at sampling point versus (a) mean hydraulic diameter of vessels (R 2 = 0.765, P < 0.001) and vessel density (R 2 = 0.467, P < 0.001), (b) theoretical specific conductivity (R 2 = 0.513, P < 0.001) and leaf-specific conductivity of xylem (R 2 = 0.618, P < 0.001)

Full size image
Fig. 5
figure 5

Reverse relationship between vessel diameter and vessel density

Full size image

The stronger relationship between d b and LSCt compared to k t (Fig. 4b) resulted from Huber values largely determined by branch diameter (Sellin and Kupper 2006). The branches of silver birch sampled in the present study contained heartwood in marginal quantities, the non-functional region in the centre of branches mostly comprised up to 4% of their total cross-sectional area (Fig. 6). Only in very thin distal parts of the branches, the non-functional area extended up to 8% due to relatively big proportion of the pith.

Fig. 6
figure 6

Branch diameter versus fraction of functional xylem of the branch cross-sectional area measured under bark

Full size image

Variation with distance from trunk

The distribution of inner diameters of the vessels varied in main branches with distance from the trunk (Fig. 7). In the proximal and middle parts of the branches, the frequency of vessels decreased smoothly from the 10–15 to 55–60 μm diameter class. However, the largest contribution to water transport in the branch base were from the 35–45 μm vessels and in the branch centre from the 30–40 μm diameter vessels. In the distal part of the branch, the modal size class (24.0% of all vessels) was unexpectedly wider, 20–25 μm, while the vessel frequency decreased steeply with increasing diameter (Fig. 7c); most of the xylem water transfer was attributable to the vessels 25–35 μm in diameter.

Fig. 7
figure 7

Frequency distributions of vessel lumen diameters and the contribution of each diameter class to hydraulic conductivity of xylem in the proximal (a), middle (b) and distal (c) parts of the branches. Mean vessel diameter (D) in the distal part differed from that both in the middle (P < 0.05) and proximal parts (P < 0.001) of the branch; the proximal and middle parts did not differ in D. σ 2, variance

Full size image

Distance from the trunk had a significant effect on vessel hydraulic diameter, vessel density and vessel relative area (Table 2). D h decreased from an average of 37.6 μm at the branch base to 29.4 μm in the distal part of the branches, while VD and VA, in contrast, increased in branch distal direction (Fig. 1). However, according to the Tukey test, the differences in the mean VA along the main branches were insignificant. Neither k t nor LSCt depended on distance from the trunk when covariates were included in the ANCOVA model (Table 2). After the covariates had been removed, the effect of sample location on the theoretical hydraulic characteristics of xylem became highly significant (P < 0.001). The covariates, d b in particular, explained the major part of the variation in both characteristics (Table 2, Fig. 4b), accounted for the factors included in the final models. Thus, the decrease in k t and LSCt in the branch distal direction (Fig. 2) is primarily associated with the branch taper.

Table 2 Results of ANCOVA for the effects of the distance from trunk and other factors on anatomical and hydraulic characteristics in main branches (sampling locations Pr, Mi and Di)
Full size table

Variation with branch order

Maximum lumen diameter of vessels extended up to 60, 50 and 45 μm in the first-, second- and third-order branches, respectively (Fig. 8). The modal diameter class was 10–15 μm for the main branches and 15–20 μm for higher-order branches. The largest contribution to hydraulic conductivity was made by vessels of 35–40, 25–30 and 20–25 μm in the I, II and III order branches, respectively. D h depended significantly (P = 0.011) on branching hierarchy, declining steeply with increasing branch order (Fig. 1a). Vessel density and vessel relative area were independent of branch order when covariates were included in the ANCOVA model (Table 3). The effect of branch order on VD became highly significant (P < 0.001) after the removal of the covariates. Both k t and LSCt decreased significantly (P < 0.01) with increasing branch order (Fig. 2). The highest specific conductivity of xylem (5.15 ± 0.405 kg m−1 s−1 MPa−1; mean ± SE) was observed in the proximal part of the main branches situated in the midcrown and the highest leaf-specific conductivity (1.36 × 10−3 ± 0.26 × 10−3 kg m−1 s−1 MPa−1) in the proximal part of the main branches situated in the lower crown. Both hydraulic characteristics were lowest in the tertiary branches: k t = 2.11 ± 0.112 kg m−1 s−1 MPa−1; LSCt = 2.74 × 10−4 ± 0.24 × 10−4 kg m−1 s−1 MPa−1.

Fig. 8
figure 8

Frequency distributions of vessel lumen diameters and the contribution of each diameter class to hydraulic conductivity of xylem in the first- (a), second- (b) and third-order branches (c). All three branch orders differed (P < 0.01) from each other in mean vessel diameter (D). σ 2, variance

Full size image
Table 3 Results of ANCOVA for the effects of branch order and other factors on anatomical and hydraulic characteristics of the branchwood (sampling locations I, II and III)
Full size table

Discussion

Vertical patterns within a crown

Mean diameter of vessel lumina in branchwood of B. pendula for three canopy layers was uniform, varying merely from 22.6 to 23.3 μm, coinciding with the mean value (23.0 μm) based on both branch- and stemwood samples analysed by Hellberg and Carcaillet (2003). As a consequence of the lack of variation in D, neither D h nor k t exhibited a vertical trend within the crown of silver birch (Figs. 1a, 2a). Thus, the substantial differences in the apparent soil-to-leaf conductance (G T) observed between different crown layers (Sellin and Kupper 2005a, b) cannot result from variation of conducting efficiency of the branchwood (i.e., vessel size distribution). As Huber values decreased acropetally within a birch canopy (Sellin and Kupper 2006), LSCt even increased in the direction of crown base (Fig. 2b). Hence, LSCt is governed mainly by HV (Pearson r = 0.829, P < 0.001), suggesting that changes in carbon allocation patterns, rather than changes in xylem anatomy, drive the adjustment of leaf-specific hydraulic efficiency within crowns of silver birches. This is not surprising, as both intra- and interspecific studies have shown that regulation of HV is a primary means for maintaining a balance between gas- and liquid-phase conductances in response to environmental variation (Mencuccini and Grace 1995; DeLucia et al. 2000; Maherali and DeLucia 2000; Bucci et al. 2005; Edwards 2006). However, Protz et al. (2000) published contrary results on Pinus contorta Dougl. ex Loud., suggesting that the principle hydraulic adjustment within a crown involves changes in sapwood conductivity rather than in HV.

The results do not support our hypothesis on reasons for vertical variation of G T in silver birch, and contradict several other studies (e.g., Lemoine et al. 2002; Aasamaa et al. 2004; Jerez et al. 2004; Burgess et al. 2006) establishing acropetal increase in both specific and leaf-specific hydraulic conductivity of branchwood within tree crowns. The differences in G T with respect to crown position in silver birch probably ensued from two factors. First, the length of the water transport path within branches was not even throughout the crown. In fact, the lower-crown main branches were on average 10.5% longer than the upper-crown branches, although the difference was statistically insignificant because of the high variation in branch length within the crown layer (Sellin and Kupper 2006). Warren and Adams (2000) confirmed a strong (R 2 = 0.842) negative exponential relationship between branch length and G T in Pinus pinaster Ait. In Pinus sylvestris, whole-branch hydraulic conductance (G B) was related positively to branch diameter and negatively to branch vertical position (Mencuccini and Grace 1996). This resulted in lower G B for the branches at the crown bottom than branches of the same diameter at the treetop.

One must take into account that the older and longer branches situated at the base of the crown bear considerably longer side branches and contain more nodes/branch junctions, i.e., regions with low xylem conductivity (Rust and Hüttl 1999; Tyree and Zimmermann 2002; Schulte and Brooks 2003). In Fraxinus excelsior L., both whole-branch hydraulic resistance (R B = 1/G B) and leaf-specific resistance increased with increasing distance from the branch to the treetop (see Fig. 7 in Cochard et al. 1997). The development of secondary axes containing short internodes was primarily responsible for the increase in R B. Growing hydraulic constraints on leaf water supply resulting from longer path length within branches have been proved in several tree species using a carbon isotope discrimination analysis (Warren and Adams 2000; Brendel et al. 2003; Samuelson et al. 2003).

The second reason for the spatial patterns of G T within a crown of silver birch might be leaf hydraulic conductance, which depends largely on light availability. Leaves account for a majority (50–90%) of the hydraulic resistance to water flow through shoots of trees (Yang and Tyree 1994; Nardini and Salleo 2000; Nardini 2001), and thus form a substantial part of the whole resistance of the water transport pathway from soil to leaves (Sack et al. 2003; Sack and Holbrook 2006; Sellin and Kupper 2007a). However, not in all cases, majority of the total hydraulic resistance occurs at the leaf level (Gyenge et al. 2005). Hydraulic conductance of sun leaves in a number of temperate woody species has been found to be significantly higher than that of shade leaves (Sack et al. 2003; Sellin and Kupper 2007a, b). The same trend holds true for interspecific comparisons: maximum leaf hydraulic conductance in both shade-growing pteridophytes and angiosperms was much lower than that in the sun-dwelling species (Brodribb and Holbrook 2004).

Patterns with respect to the distance from trunk

The anatomical as well as hydraulic characteristics exhibited pronounced trends with respect to the distance from the trunk: D h, k t and LSCt decreased, whereas VD and VA increased in the branch distal direction (Figs. 1, 2). A similar trend has been described for other woody species: contribution of the branch segments to the whole-branch resistance increased exponentially from the base to the tip of the branch (Yang and Tyree 1993; Mencuccini and Grace 1996). As both hydraulic characteristics in silver birch depended on the branch diameter measured under bark (Fig. 4b), their trends with distance from the trunk were related largely to the branch taper (Table 2). To provide adequate water supply to foliage, xylem conductivity must increase with increasing d b, because branches of bigger diameter were longer and sustained larger transpiring area. Therefore, a positive relationship between stem or branch diameter and hydraulic conductivity is a universal principle recognised in dozens of species from different systematic groups (Coyea and Margolis 1992; Yang and Tyree 1994; Mencuccini and Grace 1996; Maherali and DeLucia 2000; Tyree and Zimmermann 2002).

However, the spatial patterns observed in this study cannot be explained merely with long-distance water transport, because crown architecture needs to satisfy both hydraulic and mechanical demands. The increase in VD and VA in the branch distal direction (i.e., increasing proportion of fibres in the proximal direction) provides greater mechanical strength to resist the greater physical load befalling to the branch base due to the longer force arm and its greater weight including the cumulative weight of laterals and leaves. The low vessel density in the branch bases is functionally balanced by their high conducting efficiency, i.e., by large D h (Figs. 1a, b). Thus, the strong negative relationship revealed between vessel diameter and density (Fig. 5) probably implies a basic trade-off between hydraulic efficiency and mechanical stability of xylem. A study in congeneric chaparral species (Wagner et al. 1998) has confirmed substantial impact of the size and frequency of vessels within stems on their mechanical properties. Vessel lumen relative area could even override the importance of variation in fibre anatomy and lignification.

In silver birch, D h and VD combined explained 85.4% of the total variation in specific conductivity of xylem in the whole data set. The negative relationship between vessel diameter and density seems to be a more general feature of xylem anatomy, reflecting optimisation of wood structure to meet diverse functional demands (mechanical support, water transport, respiration of living tissues, etc.) at both ecological and evolutionary scales. The numerical data published by Hellberg and Carcaillet (2003) manifest a close correlation (R 2 = 0.974, P < 0.05) between D and VD across four European species of Betula. The opposite trends in D and VD have also been observed in other studies both within (Mokany et al. 2003) and across species (Vander Willigen et al. 2000; Preston et al. 2006). As a result of this negative relationship, mean VA varied within a narrow range (14.0–16.2%) along main branches of silver birch. VA was determined by VD rather than by D, unlike that published by Wagner et al. (1998) for chaparral shrubs. In Eucalyptus camaldulensis (Dehn.), shifts in both VD and D were responsible for the variation in VA as a result of environmental changes (Thomas et al. 2004).

Patterns with branch orders

Four of the five variables depended significantly on branch ranking: D h, k t and LSCt decreased, whereas VD increased with increasing branch orders (Figs. 1, 2). Primary branches had on average 2.1-fold higher k t compared with third-order branches, primarily due to differences in the frequency distribution of vessel diameter (Fig. 8). Thus, the water transport capacity declined from main branches to branches of higher orders. The pronounced hydraulic hierarchy revealed among the branches may have strong implications on distribution of the transpiration stream within a crown. This pattern promotes the hydraulic dominance of primary branches over secondary branches and their dominance over the tertiary branches; these results are consistent with the hydraulic segmentation hypothesis (Tyree and Zimmermann 2002). Thus, differentiation of branches by hydraulic capacity with branch orders may be interpreted as a mechanism by which crown architecture contributes to preferential water flow along the main axes, potentially providing better water supply to the branch apical bud and foliage located in the outer, better-insolated part of the crown.

Specific conductivity of xylem in silver birch depended significantly (Tables 1, 2, 3) on growth ring width (W): the greater the branch radial growth rate, the higher the water conducting efficiency of the xylem to be formed. Positive correlation between xylem hydraulic conductivity or conduit diameter and stem growth rate has been proved both in gymnosperm and in angiosperm species (Pothier et al. 1989a; Medhurst and Beadle 2002; Reich et al. 2003; Cochard et al. 2005). Because neither D nor D h was correlated with W in silver birch, higher k t observed in the faster-growing branches/zones (primary branches, branch middle parts) is rather attributable to larger relative area of vessel lumina. In the analysis across different species with diverse wood anatomy (coniferous, diffuse- and ring-porous trees) an opposite trend has been observed (McCulloh and Sperry 2005). According to the results of McCulloh and Sperry (2005), the trend toward greater hydraulic efficiency in trees is associated with a decrease in sapwood area occupied by xylem conduits. In the case of B. pendula, a heteroxylous angiosperm species, VA was statistically independent of vessel diameter (Pearson r = 0.098, n.s.), in contrast with homoxylous wood of conifers, in which the percentage of sapwood area occupied by tracheid lumina is proportional to tracheids’ mean diameter (Pothier et al. 1989b).

In terms of the mean LSCt, the primary and third-order branches differed by a factor of 2.5, attributable to the variation in both HV and k t. As HV in silver birch also exhibited a trend to decrease with increasing branch orders (see Fig. 1b in Sellin and Kupper 2006), a positive correlation was revealed between k t and HV. This is contrary to the data on hydraulic properties across multiple tree species, different growth conditions or seasons (Vander Willigen et al. 2000; Macinnis-Ng et al. 2004), demonstrating reverse trends in the two characteristics. In this way, low specific conductivity of xylem is balanced by higher Huber values. This is regarded as homeostatic response to increasing hydraulic constrictions to maintain a reasonable stomatal opening across diverse habitats by adjusting sapwood area to leaf area ratio. The apparent contradiction between our data and those referred to above likely results from differences in plant requirements to meet various environments, in species growth strategies or sampling schemes. Hence, the relation of the trends in xylem specific conductivity and Huber value depends on scale (i.e., intra- vs. interspecific study, single versus multiple contrasting habitats, single vs. multiple sampling points within individuals).

To summarise, the differences in soil-to-leaf hydraulic conductance in B. pendula between the upper- and lower-canopy foliage did not result from the vertical variation in hydraulic efficiency of branchwood. The acropetal increase in G T within a birch canopy is rather attributable to the longer path length within the lower-crown branches and higher hydraulic resistance of the shade leaves.