1. Plants must balance water expenditure from their crown with water supplied through root and stem tissues. Although many different combinations of hydraulic traits could accomplish water balance, we ask whether variation across species in stem hydraulic traits has been concentrated along few, or many, dimensions of trait variation.
2. We measured stem hydraulic traits for 120 woody dicot species across a range of different biomes in eastern Australia. Mean annual temperatures ranged from 10 to 27 °C and aridity (precipitation/potential evapotranspiration) from 0·33 to 1·02 across study sites.
3. Xylem-specific conductivity, species’ height and ratio of leaf area to xylem area were positively correlated, manifesting as a single axis of trait variation, with other traits mostly orthogonal to this axis. Thus, as height and ratio of leaf area to xylem area increased across species and habitats (increasing resistance per leaf area), xylem-specific conductivity partially compensated for this resistance. Xylem-specific conductivity was well predicted by increasing height (r2 = 0·45) and ratio of leaf area to xylem area (r2 = 0·36). This three-trait axis was positively correlated with increasing precipitation (r2 = 0·28) and temperature (r2 = 0·15), but most of the explained variance lay within sites (39%) rather than across sites (10%). Thus, the spread of species’ traits along this functional axis reflected structural and hydraulic differences among co-occurring species, at least as much as it reflected differences associated with contrasting climates.
4. High xylem-specific conductivity in stems was accomplished by high vessel diameter to number ratio (r2 = 0·32) and/or by high vessel lumen fraction (r2 = 0·13). Low midday water potential (higher xylem tension) was associated with low ratio of vessel diameter to number (r2 = 0·25), whereas low specific gravity (r2 = 0·18) and stiffness (r2 = 0·12) were associated with high vessel lumen fraction.
5. Light capture (i.e. increasing height and leafiness) may be facilitated by high xylem-specific conductivity, but marked increases in xylem-specific conductivity may also be associated with reduced hydraulic and mechanical safety. Although the trade-offs associated with increasing xylem-specific conductivity remain unclear, our data suggest that xylem-specific conductivity is important for maintaining water balance across a large range of species and biomes.
Functional plant traits are confined by the physical environment, physiology, structure and biotic interactions; there are only so many combinations of plant traits that will result in a viable and competitive outcome. It is not surprising then that contemporary efforts exploring trait variation among species and habitats have discovered that a few main dimensions of trait variation explain much of the total variation across species and habitats (Grime et al. 1997; Reich, Walters & Ellsworth 1997; Westoby et al. 2002; Díaz et al. 2004; Zanne et al. 2010). Here, we ask whether a common axis of variation exists among stem hydraulic traits, and if so, what are the likely trade-offs underpinning it.
Water must be delivered from the soil to the sites of photosynthesis, mainly leaves in woody dicot species. Although conceptually this is straightforward, various aspects of the soil, atmosphere, plant structure and plant physiology influence the delivery of this water. A simple proportionality modified from the Whitehead–Jarvis application of Darcy’s law (Whitehead & Jarvis 1981) can help us understand how these various plant traits and environmental variables might relate to one another:
KS (xylem-specific conductivity) is the rate of sap-flow per cross-sectional area of xylem across a given pressure gradient (length-normalized). LA/XA is the ratio of leaf area to xylem area of a plant or shoot. The left side of the proportionality (KS/LA/XA) yields leaf-specific conductivity, conductivity per unit leaf area. Water potentials in soil and leaves are given as ΨS and ΨL, respectively. Leaf water potential, measured pre-dawn, reflects soil water potential (ΨS) and is used as a proxy measure for water availability within sites (e.g. rooting depth). The difference between soil water potential (ΨS) and leaf water potential (ΨL) approximates the pressure gradient driving plant conductance. VPD, Ht and gS denote leaf-to-air vapour pressure deficit, crown height and stomatal conductance, respectively. We address in the methods several assumptions intrinsic to this version of the proportionality. Also, when using the term ‘xylem’ throughout the text, we consider only stem tissue.
We measured hydraulic traits relating to Darcy’s law for 120 species of woody dicots across a large range of latitude and precipitation in eastern Australia. The range of climates represented by this sampling included roughly a third of the total biome space on earth as indexed using Whittaker’s biome classification scheme (Whittaker 1975) (Fig. S1, Supporting information). We ask three main questions. (i) Which of the hydraulic traits identified in Darcy’s law contribute most to variation across species and sites? (ii) How much of the variation in these traits is attributable to differences between coexisting species, and how much variation is attributable to differences between sites? We report that xylem-specific conductivity is an important trait achieving water balance across species. Considering this, (iii) how is high xylem-specific conductivity typically achieved, and consequently, what costs might be associated with it?
Materials and methods
Site and species
Sites were chosen in three bands of latitude (c. 18, 33 and 43°S) and correspondingly experienced a wide range of temperature (MAT: 10·0–26·9 °C) (Table S1, Fig. S2, Supporting information). Within each latitude, two or three sites of contrasting aridity (precipitation/potential evapotranspiration) were selected for study. At each latitude, a ‘Wet’ site was selected where rainfall was nearly equal to potential evapotranspiration. ‘Dry’ and ‘Arid’ sites were placed where rainfall was about 60% and 35% of potential evapotranspiration, respectively (Table S1, supporting information). An arid climate by this definition was not available at the highest latitude (Tasmania). Average leaf-area index (LAI) values for each site (1 km2 resolution) were calculated from 8-day values downloaded from MODIS (2011) for the periods between December 2008 and February 2009.
Sites represented late successional vegetation on soils considered oligotrophic based on parent material, vegetation type and soil survey data, where available. Within each site, the most common woody angiosperms present (15–18 species per site, excluding vines) were chosen for analysis (Table S2, supporting information). Of the 120 species present across all sites, twelve species were encountered at more than one site. Thus, differences between sites mainly reflect intrinsic differences between species, but differences within species occurring across sites can also be assessed to some extent. Including or excluding these 12 species did not change the outcomes of our hypothesis tests. Surface soils were collected for nutrient analysis (total N, total P, organic C, cation exchange characteristics and pH) (Table S1, supporting information). Soil profiles to a depth of 1·5 m (where possible) were described, noting colour, texture, structure, redoximorphic features and root characteristics.
Fieldwork was scheduled during the growth season at each site, but the heights of the tropical dry (c. July–August) and wet (January–March) seasons were avoided. This provided trait data corresponding to favourable times of the year, rather than seasonal extremes, with the exception of the hot–wet site, which was comparatively dry when sampled. We suspect that marginally low hydraulic conductance and CO2 assimilation rates at the hot–wet site resulted from these dry conditions. Trait collection times are as follows: September 2009 (hot–wet), November 2009 (hot–dry), May 2010 (hot–arid), October 2010 (warm–wet), June 2009 (warm–dry), April 2009 (warm–arid), January 2010 (cool–wet) and March 2010 (cool–dry).
Plant selection, replication and the ratio of leaf area to xylem area
Except where noted, all plant traits were measured on five individuals of each species within each site. Replicates representative of a species’ height and vigour were selected when possible. For species >5 m in height, smaller plants (c. 5 m) were chosen to facilitate collecting. To evaluate the effect of choosing small individuals on our hypothesis tests, analyses were conducted with a complete data set (120 species), but also using a reduced data set (65 species), containing only species <5 m in height, for which trait data reflect maximum species size (Table S3, supporting information). We report only results from the full data set here because the reduced data set yielded similar results. Maximum heights for each species were taken from published descriptions, herbarium specimen labels and field observations.
For the ratio of leaf area to xylem area, we measured xylem diameters (inside bark) along the main stem of vertical leading shoots at 5, 10, 20, 40, 80 and 120 cm from the shoot tip. The inside-bark diameter (including all woody tissue beneath the cambium) at the distal and proximal ends were measured using digital callipers (precision = 0·01 mm). Stems and leaves were separated, dried to constant mass at 70 °C and weighed (precision = 0·0001 g). Ordinary least squares regression was then used to predict leaf mass as a function of inside-bark stem diameter. Leaf mass per unit area values (methods described elsewhere) were then used to convert dry leaf mass into green leaf area for any given inside-bark stem diameter. Ratio of leaf area to xylem area was calculated in this way for 0·5, 1·0 and 2·0 cm xylem diameters, excluding pith. These three ratios of leaf area to xylem area yielded similar results, and only values of ratio of leaf area to xylem area at 1·0 cm xylem diameters are reported here.
Shoot hydraulic conductivity
Shoots 120 cm in length were cut from individuals as previously described. Large side branches were cut c. 10 cm from the main stem (in air), immediately placed in sealed heavy plastic bags and transported to the laboratory within 4 days of collection. Shoots were then rehydrated by submerging the entire shoot in water for 24 h. All remaining side branches were removed (in air) and blotted dry with a paper towel, and the exposed xylem was sealed (Loctite 550 adhesive, 779 activator; Henkel, Kilsyth, Australia). The distal diameter and total length were trimmed to 0·5 mm and 80 cm, respectively. Branch conductivity was measured in the distal-to-proximal direction (Mitchell et al. 2008) using a Sperry apparatus (Sperry, Donnelly & Tyree 1988) after trimming both ends under water. We measured Kh in the distal-to-proximal direction to reduce flow through the regions of xylem associated with the sealed side branches. However, it is likely that the direction of flow has little or no effect on Kh (John Sperry, personal communication).
Branch conductivity was measured across a 100-kPa pressure gradient using 0·02 m KCl (Zwieniecki, Melcher & Holbrook 2001) that had been filtered (0·2 μm) and degassed under vacuum. Conductivity was measured at high pressure (100 kPa) to flush native emboli (Choat, Sack & Holbrook 2007) as well as emboli introduced during branch trimming. Pressure was provided using an N2 cylinder. Water was collected and weighed to the nearest 0·00001 g (Sartorius CP225D, Göttingen, Germany) and logged at 15-s intervals to calculate flow rate. To prevent contact between the N2 and the hydraulic solution (to avoid gassing the solution), the N2 was allowed to flow into a rigid polystyrene bottle that contained a flexible polypropylene bag inside it, which contained the hydraulic solution. As N2 flowed into the rigid bottle, 100 kPa of pressure was exerted on the polypropylene bag and pushed the hydraulic solution through the stem segment.
Steady-state flow rates were typically achieved after 10–20 min, although occasionally shoots required an hour or more to reach steady-state. After branch conductivity measurements, each segment was tested for open vessels by pressurizing the distal end (100 kPa N2) and recording whether bubbles were observed from the submerged proximal end. Although the majority of species had closed vessels during Kh measurement, 26 species did not. These species did not exhibit significantly higher branch conductivities than closed-vessel species (Fig. S3, supporting information), nor did excluding these species from our analysis change the outcomes of our results. Therefore, we report analyses that included all species. Xylem-specific conductivity was calculated by dividing branch conductivity by the distal cross-sectional area (xylem) of each segment, subtracting pith.
Pre-dawn (Ψpd) and midday (Ψmd) leaf water potentials were measured on two samples from each plant using a pressure chamber (Model 1000; PMS Instruments, Corvallis, OR, USA). Leaves or small twigs from the uppermost crown were cut from each plant, sealed inside a plastic bag with a damp paper towel and transferred to the pressure chamber for measurement. An effort was made to collect midday potentials during hot, sunny days.
Seven-step light response curves (PFD = 1500–0 μmol m−2 s−1) were constructed in situ for the first plant of each species at each site to determine the appropriate light saturation intensities at which to measure maximum assimilation (Amax) and transpiration (E) rates using a portable photosynthesis system (Model 6400xt; LI-COR Biosciences, Lincoln, NE, USA). Assimilation rates at each step were judged stable when varying less than c. 0·5 μmol m−2 s−1 for 5 min. Curves were visually inspected in the field to determine the appropriate light intensity for Amax measurements. Measurements were taken between 08:00 and 11:30. Within-chamber CO2 concentration, temperature and VPD were kept between 388 and 402 ppm, 23·0–27·0 °C and 1·9–2·1 kPa for all measurements. Although 2·0 kPa VPD results in reduced stomatal conductance, we note that CO2 assimilation is relatively flat between 0·4 and 2·0 kPa, even for rain forest plants (Cunningham 2004, 2005), and therefore, our across-species comparisons of Amax should remain relatively unbiased. We note that leaf transpiration rates of our wet forest species may have plateaued or even declined at 2 kPa VPD (Oren et al. 1999; Cunningham 2004) and should be interpreted with caution. Leaf mass per unit area (LMA) was calculated by dividing dry leaf mass by fresh leaf surface area.
Vessel and xylem property traits
Vessel anatomy was sampled from a single stem cross section from each species (n = 1). However, within-species replication (3) was obtained for 24 species, and mean standard deviations are reported for these species (Table 2). One branch segment (c. 0·5 cm diameter) from each species from each site was sectioned with a sledge microtome (Reichert, Vienna, Austria), mounted on a glass slide in a glycerol drop and examined under 200× magnification using a light microscope (Model BH-2; Olympus, Tokyo, Japan). A grid (484 intersections mm−2) was superimposed over the magnified cross section. Intersections were counted in three different areas within one cross section (121 intersections per count), making note of whether intersections fell on vessel lumen or other cell structures. Total vessel lumen fraction was calculated as the percentage of total intersections that fell within vessel lumen. Vessel density (number of vessels mm−2, N) was measured by counting all vessels within a superimposed 0·25 mm2 square (under 200× magnification) and dividing the number of vessels by 0·25. Mean vessel density was calculated as the average for three separate areas within one cross section per species. Mean vessel area () was calculated by dividing vessel lumen fraction by vessel density.
Xylem-specific conductivity mainly increases either by (i) increasing vessel lumen fraction (the total cross-sectional area used for sap transport, ) or by (ii) increasing the diameter of vessels within a constant vessel lumen fraction, that is, by changing the ratio of vessel area to vessel number (). Mechanism (ii) is effective because the conductivity of a conduit increases to the fourth power of its diameter; thus, conductivity may increase with increasing vessel width even if the number of vessels decreases. This is a useful way to think about vessel anatomy because vessel lumen fraction and S are orthogonal to one another and may be associated with different costs (Zanne et al. 2010). For example, if species achieve conductivity via vessel lumen fraction, specific gravity and strength may decrease as a result. However, there is no reason to expect conductivity and xylem strength to be inversely correlated if conductivity is achieved via increasing S (i.e. increasing Ā/N without increasing vessel lumen fraction ).
Specific gravity was calculated as fresh volume divided by oven-dried mass. Drying was carried out at 70 °C. Further drying to 103 °C for of a subset of specimens changed their mass by less than one per cent. Volume measurement was preceded by submersion in water overnight. Volume was estimated on xylem segments (bark and pith removed) by submerging a suspended stem segment in a beaker of water that had been placed on a laboratory precision balance. The displaced mass (i.e. volume) of water was then recorded.
After measuring xylem-specific conductivity on the segments described previously, modulus of elasticity (aka Young’s modulus), a measure of material stiffness, was measured on these identical segments. A general materials testing machine (Model 5542; Instron Corporation, Canton, MA, USA) was used to perform a three-point bending test, ensuring the length-to-width ratio of segments exceeded 20.
Darcy’s law assumptions
The Darcy’s law proportionality given in the introduction has several embedded assumptions. The first is that xylem conductivity measured at one point along the path length (in this case, in terminal branch xylem) reflects differences in conductivity summed across the entire path length. Although this assumption may not necessarily hold for all species, conductivity measured on individual plant organs (e.g. branches) and that on whole plants are often closely correlated (Nardini & Salleo 2000; Brodribb, Holbrook & Gutiérrez 2002; Fichot et al. 2011). Secondly, we assume that hydraulic resistance increases proportionately with height or path length, which is also unlikely to hold true within or across species (Yang & Tyree 1994; Becker, Tyree & Tsuda 1999). Third, we have omitted water viscosity and gravity because they contribute little to variation across species (although viscosity does decrease slightly with site temperature). Given these assumptions, the Darcy’s law proportionality is a useful guide for interpreting correlations between hydraulic traits but is not intended as an accurate quantitative equation.
Data and phylogenetic analysis
Ordinary least squares (OLS) regression was used to measure bivariate and multivariate relationships among plant traits and climate variables. Log10-transformed data were used in these analyses (except where noted), as data were generally right-skewed. Principal component analysis (PCA) was used as a dimension reduction technique for evaluating relationships between multiple traits. We also used PCA to examine more closely the relationships between the major axis of variation and other plant traits and climate variables. To do this, we included only the main-axis traits (e.g. xylem-specific conductivity, height and ratio of leaf area to xylem area) in the PCA and calculated the species scores on the first principal component. We then used these species scores in further bivariate analyses to measure the relationship between the major axis of variation and other variables of interest such as pre-dawn leaf water potential. Data were standardized by subtracting the mean and dividing by the standard deviation prior to PCA.
To assess the contribution of within-site correlations to r2 values, we wrote a separate sampling procedure in R (R Foundation for Statistical Computing, Vienna, Austria). Preliminary analyses suggested that KS, height and LA/XA were the most variable and strongly correlated hydraulic traits in the data set. Thus, further analyses were performed to explore the possible functional relationships between these three traits. According to Darcy’s law, KS should increase proportionately to the product of height and LA/XA. An ordinary least squares regression was run on these two traits of interest (e.g. KS = height × LA/XA), using mean species values. We then shuffled the x (height × LA/XA) and y (height) values independently within each site (eight sites), keeping the original values but randomizing their order. Thus, the means and variance in x and y within sites remained unchanged, but any correlation between them was removed. We then ran the regression again, this time using the shuffled data set and repeated this procedure 1000 times, taking the mean from the generated r2 distribution. We interpret this r2 value as the percentage variance explained across sites, having completely removed any within-site correlation. The within-site contribution to the total r2 was then calculated by subtracting the r2 from the shuffled data set from the non-shuffled data set r2.
We generated a phylogenetic tree for our species using the Phylomatic Project website (Webb & Donoghue 2011) and the ‘maximally resolved seed plant tree’ (Stevens 2011). Polytomies were resolved where possible from other published sources. We did not assign branch lengths to our taxa because branch lengths are not known for most of our genera and species. Output from the Phylomatic Project website was saved as a Newick file, which was then brought into the Picante package (Kembel et al. 2010) for R (version 2.12.1, R Development Core Team 2010). Phylogenetically independent contrasts (PICs) were estimated for key traits using the ‘pic3’ function in Picante (Table S4, supporting information). The role of deep compared to recent evolutionary divergences was assessed by comparing correlation coefficients using PICs vs. raw data (i.e. across present-day species).
Which traits accounted for the most variation across species and sites?
Xylem-specific conductivity (KS) measured near the tip of the shoot was positively correlated with the maximum attainable height of a species (r2 = 0·45; P < 0·001) and with the ratio of leaf area to xylem area (r2 = 0·35; P < 0·001) (Table 1). As suggested by Darcy’s law, xylem-specific conductivity was well predicted (r2 = 0·49, P < 0·001) by the product of plant height and the ratio of leaf area to xylem area (Fig. 1). Xylem-specific conductivity, height and the ratio of leaf area to xylem area formed a major axis of variation across all species and sites, with leaf water potentials and leaf transpiration largely falling orthogonal to this axis (Fig. 2).
Table 1. Pearson product moment correlation matrix for mean species values across all sites (n = 8)
Correlation coefficients (r) and associated P values (beneath r values) are shown for all hydraulic trait correlations.
MAP, mean annual precipitation; MAT, mean annual temperature; KS, xylem-specific conductivity (n = 5); Ht, maximum plant height; LA/XA, leaf-area/xylem-area ratio (n = 5); Ψpd, leaf water potential at pre-dawn (n = 5); Ψmd, leaf water potential at midday (n = 5); E, leaf transpiration (n = 5); VLF, vessel lumen fraction (n = 1); S, vessel area/vessel density (n = 1); KLA, leaf-specific conductivity (n = 5).
Further examination (see PCA analysis methods) revealed that this main axis of variation was weakly, but positively correlated with leaf-level transpiration (r2 = 0·13, P < 0·001), Amax (r2 = 0·09, P = 0·001), pre-dawn leaf water potential (r2 = 0·22, P < 0·001) and midday leaf water potential (r2 = 0·17, P < 0·001). In addition to the variables in Darcy’s Law, the three-trait axis was positively correlated with leaf size (r2 = 0·56, P < 0·001), vessel diameter (r2 = 0·49, P < 0·001), vessel lumen fraction (r2 = 0·11, P < 0·001), S (r2 = 0·44, P < 0·001), precipitation (r2 = 0·28, P < 0·001) and temperature (r2 = 0·15, P < 0·001) and was negatively correlated with specific gravity (r2 = 0·17, P < 0·001), stem modulus of elasticity (r2 = 0·10, P < 0·001) and leaf-area index (r2 = 0·12; P < 0·001).
How important was within-site variation relative to across-site variation?
The components of the major three-trait axis described above (xylem-specific conductivity, height and the ratio of leaf area to xylem area) all increased along the climate gradients from cold and dry to hot and wet, but variation within sites was substantial (Table 2). For predicting xylem-specific conductivity from the product of height and the ratio of leaf area to xylem area (Fig. 1), differences between co-occurring species (within-site variance) accounted for 39% of the predictive capacity, whereas differences between sites accounted for only 10% (Fig. 1). Precipitation and temperature together in multiple regression accounted for 29%, 22% and 25% of the variance in xylem-specific conductivity, height and the ratio of leaf area to xylem area, respectively.
Table 2. Average within-species (n = 5), within-site (n = 15) and across-site (n = 8) standard deviations for all hydraulic traits
KLA, leaf-specific conductivity (mg m−1 MPa−1 s−1); KS, xylem-specific conductivity (kg m−1 MPa−1 s−1); LA/XA, leaf area (m2)/xylem area (cm2); Ht, maximum species’ height (m); Ψpd, leaf water potential measured pre-dawn (MPa); Ψmd, leaf water potential measured midday (MPa); E, leaf transpiration rate (mmol H20 m−2 s−1) measured at 2·0 kPa VPD; Vdia, mean vessel diameter (μm); Vden, vessel density (number mm−2); SG, Specific gravity (g cm−3); LMA, leaf mass (g)/leaf area (m2); LS, leaf size (cm2).
*LA/XA means were interpolated from species-specific allometric equations and therefore within-species SD was not calculated.
†Height is the maximum attainable height of the species as taken from previously published books, reports, and journal articles. As such, within-species SD is unknown.
‡Within-species replication for vessel diameter, vessel density, and specific gravity was three. For vessel diameter and vessel density, within-species replication was obtained for only 24 species (mean SD reported above).
How was high xylem-specific conductivity achieved?
Increasing xylem-specific conductivity arose through increasing S (wider, but fewer vessels) and through increasing vessel lumen fraction (Fig. S4, supporting informaiton), but the influence of S was stronger (r2 = 0·32; P < 0·001 vs. r2 = 0·13; P < 0·001). Increases in xylem-specific conductivity via vessel lumen fraction were associated with decreases in specific gravity (VLF∼SG: r2 = 0·18; P < 0·001) and tissue stiffness (modulus of elasticity, MOE) (r2 = 0·12; P < 0·001), whereas increases in xylem-specific conductivity via S were not associated with reduced tissue stiffness (Fig. 3), even though higher S was associated with decreased specific gravity (S∼SG: r2 = 0·15; P < 0·001). Increases in xylem-specific conductivity via S were also associated with higher midday leaf water potential (r2 = 0·25; P < 0·001), but only after removing species from cold and wet sites where low midday xylem potentials are unlikely to occur (Fig. 4) (Table S1, Supporting information).
Correlations between the hydraulic traits using phylogenetically independent contrasts (Table S4, supporting information) were similar to correlations across present-day species (Table 1). Bivariate correlation coefficients between leaf transpiration and the three-trait axis (sapwood-specific conductivity, height, ratio of leaf area to sapwood area) were often significantly lower when using phylogenetically independent contrasts (Table S4, supporting information). This indicates that divergence in these traits has been phylogenetically conserved to some extent. We note that this result does not indicate that the relationship between leaf transpiration and the three-trait axis is less meaningful, but only that it has arisen from earlier divergences as well as from ones that are more recent (Westoby, Leishman & Lord 1995).
Which traits accounted for the most variation across species and sites?
Xylem-specific conductivity appeared to be a key trait achieving water balance (reducing whole-plant resistance) as path-length resistance and ratio of leaf area to xylem area increased (increasing demand) across species. Although increasing access to soil water (as indexed by pre-dawn leaf water potential) also served to offset increasing height and ratio of leaf area to xylem area, this mechanism was mostly evident across climate gradients. Transpiration rates did not decrease with plant height as would be expected if nothing else had changed in Darcy’s law. Rather, transpiration increased with increasing height and ratio of leaf area to xylem area. Thus, xylem-specific conductivity, and to some extent soil water potential (e.g. increasing rooting depth or decreasing aridity), were the only plant traits (viz Darcy’s law) adjusted to counter increasing resistance (i.e. taller plants) and increasing water demand (i.e. leafier shoots) across the species and sites. This suggests that previous findings of decreasing transpiration and ratio of leaf area to xylem area with increasing height (McDowell et al. 2002; Martínez-Vilalta et al. 2009) were largely a within-species phenomenon and that across species increasing height is more generally associated with increasing ratio of leaf area to xylem area and xylem-specific conductivity, even within the same site. However, we note that the maximum xylem-specific conductivity values of our species are near the maximum recorded world-wide (B. Choat et al. unpublished data), and it is unknown whether xylem-specific conductivity would continue to compensate for increasing height beyond the maximum values presented here (c. 35 m).
How important was within-site variation relative to across-site variation?
Although xylem-specific conductivity increased significantly with precipitation and temperature, the majority of covariance between this trait and height and ratio of leaf area to xylem area was explained within sites, not across climate gradients. This suggests that the location of species along this functional trait axis (xylem-specific conductivity, height and ratio of leaf area to xylem area ratio) does reflect environmental filtering across large scales but is an even more variable and important indicator of functional differences between species within sites.
We suggest that natural selection acting on traits conferring greater light interception (taller stature, leafier shoots and more conductive xylem) has resulted in much covariance in these traits across light gradients, where coexistence among shade-tolerant and shade-intolerant species is possible. Although height may be an outcome of other selective pressures such as escaping savanna fires (Gignoux, Clobert & Menaut 1997), height is known to have a profound influence on light availability (Osada et al. 2004), leaf C assimilation (Thomas & Bazzaz 1999) and growth rate (Gleason et al. 2009). Game theoretical models predict increasing plant height in high-competition environments where light limits growth (Iwasa, Cohen & Leon 1984; Falster & Westoby 2003). Thus, as habitats become less water-limited and more light-limited, plants must either become taller in pursuit of light or alter their structure and physiology to make a living in the shade.
How was high xylem-specific conductivity achieved?
Increasing xylem-specific conductivity can be accomplished two main ways. (i) Vessel lumen fraction (i.e. the cross-sectional xylem area used for water transport) may increase or (ii) the average vessel diameter at a given vessel lumen fraction (i.e. the ratio of vessel area to vessel number; S) may increase (Zanne et al. 2010). Because xylem-specific conductivity scales to the 4th power of vessel diameter, increasing S yields greater xylem-specific conductivity even if there are fewer vessels.
The costs or risks of increasing xylem-specific conductivity are different depending on whether the increase in xylem-specific conductivity occurs via vessel lumen fraction or via S. Increasing the fraction of xylem that is vessel lumen essentially increases the fraction of xylem that is empty space and therefore decreases specific gravity (Preston, Cornwell & DeNoyer 2006; Russo et al. 2010), a strong predictor of wood mechanical properties (Niklas & Spatz 2010). Such stems should be less resistant to bending and breaking at a given diameter than stems of higher specific gravity (Wagner, Ewers & Davis 1998; Sperry, Meinzer & McCulloch 2008). Thus, reduced xylem strength and stiffness presents a possible trade-off with increases in conductivity achieved by greater vessel lumen fraction. Vessel lumen fraction varied from 3·6% to 27·0% in this study but was a weak predictor of xylem-specific gravity. The mechanical weakening associated with increasing vessel lumen fraction might be accepted as a risk or might be offset by: (i) increasing the specific gravity of non-lumen tissues (Poorter et al. 2010; Zanne et al. 2010) or (ii) building wider stems, with little (if any) additional mass cost per length (Anten & Schieving 2010).
Increasing S accounted for 33% of variance in xylem-specific conductivity across the species in this study, whereas increasing vessel lumen fraction accounted for only 13% of the variance in xylem-specific conductivity. There was no evidence that increasing S compromised wood or branch mechanical properties (Butler et al. 2011), even though S was a weak predictor of xylem-specific gravity. However, empirical data and theory suggest that wide vessels, associated with increasing S, are more vulnerable to embolism risk (Zimmermann 1983; Tyree, Davis & Cochard 1994; Jacobsen et al. 2007). Therefore, it is reasonable to expect an evolutionary trade-off between xylem-specific conductivity and embolism resistance. Although some data contradict this view (Maherali, Pockman & Jackson 2004; Bhaskar, Valiente-Banuet & Ackerly 2007), other studies (Hacke et al. 2006; Pittermann et al. 2006; Fichot et al. 2011; Markesteijn et al. 2011) and the results reported here support it. It is likely that the complex relationship between vessel diameter and embolism risk (Hacke et al. 2006) and marked differences between species and microhabitats have thwarted some efforts to reveal this trade-off. For example, embolism resistance is not likely to increase the fitness of narrow-vessel species occurring in cool, wet habitats. After removing these species from our analysis, S and xylem-specific conductivity were strongly correlated with minimum leaf water potential. Although minimum leaf water potential is not a direct measure of embolism resistance, it is an indicator of tolerable xylem tensions prior to stomatal closure and is often closely correlated with embolism resistance (Bhaskar & Ackerly 2006; Choat, Sack & Holbrook 2007; Jacobsen et al. 2007).
Xylem-specific conductivity appears to be a key stem hydraulic trait, enhancing water supply as path-length resistance (i.e. height) and demand (ratio of leaf area to xylem area) increase across species. The likely trade-offs in reduced xylem strength and embolism safety associated with increasing xylem-specific conductivity have been pointed out previously (Zimmermann 1983; Wagner, Ewers & Davis 1998). While these ideas have attracted less attention in recent years, they should not be forgotten.
Although other studies have demonstrated increasing xylem-specific conductivity with increasing plant height (Mokany et al. 2003; Calvo-Alvarado, McDowell & Waring 2008; Poorter et al. 2010; Zach et al. 2010), the extent of this relationship across diverse habitats and species was unknown. We suggest that the relationship between xylem-specific conductivity, height and ratio of leaf area to xylem area is indicative of life-history strategies geared to maximize access to light (Iwasa, Cohen & Leon 1984; Falster & Westoby 2003). It is likely that warm, wet and productive environments put a premium on strategies that pursue height and light interception and that correlations between height, leafiness and xylem-specific conductivity may be underpinned by trade-offs with wood strength, and embolism resistance.
We are very grateful to the many individuals who have contributed to this study. CSIRO scientists Maria Ottenschlaeger, David Drew, Phil Smethurst and Tony Grice lent us equipment and office space. Daniel Falster collaborated on the variance partitioning routine and Drew Allen assisted with the statistical analysis. Ian Davidson assisted in the field and laboratory. Ian Wright, Peter Reich and Tanja Lenz contributed trait data. Amy Zanne, Yusuke Onoda and Hendrik Poorter assisted with xylem conductivity, materials testing and image analysis, respectively. Thanks are also given to Judy and Eddie Howitt, John and Jill Bignell, and National Parks for their hospitality and assistance with site access. We also thank Chris Blackman, Lawren Sack and two anonymous reviewers for constructive comments on earlier versions of this manuscript. This research was funded by a Discovery Project grant awarded to Mark Westoby from the Australian Research Council.