Hydraulic efficiency and safety of branch xylem increases with height in Sequoia sempervirens (D. Don) crowns



    Corresponding author
    1. School of Plant Biology, University of Western Australia, 35 Stirling Highway, Crawley WA 6009 Australia,
    2. Department of Integrative Biology, University of California, Berkeley, CA 94720, USA
    Search for more papers by this author

    1. Department of Biology 257 South 1400 East University of Utah Salt Lake City, UT 84112, USA, and
    Search for more papers by this author

    1. Department of Integrative Biology, University of California, Berkeley, CA 94720, USA
    Search for more papers by this author

Stephen S. O. Burgess. Tel./Fax: +61 (8) 6488 2073; e-mail: ssb@cyllene.uwa.edu.au


The hydraulic limitation hypothesis of Ryan & Yoder (1997, Bioscience 47, 235–242) suggests that water supply to leaves becomes increasingly difficult with increasing tree height. Within the bounds of this hypothesis, we conjectured that the vertical hydrostatic gradient which gravity generates on the water column in tall trees would cause a progressive increase in xylem ‘safety’ (increased resistance to embolism and implosion) and a concomitant decrease in xylem ‘efficiency’ (decreased hydraulic conductivity). We based this idea on the historically recognized concept of a safety–efficiency trade-off in xylem function, and tested it by measuring xylem conductivity and vulnerability to embolism of Sequoia sempervirens branches collected at a range of heights. Measurements of resistance of branch xylem to embolism did indeed show an increase in ‘safety’ with height. However, the expected decrease in xylem ‘efficiency’ was not observed. Instead, sapwood-specific hydraulic conductivities (Ks) of branches increased slightly, while leaf-specific hydraulic conductivities increased dramatically, with height. The latter could be largely explained by strong vertical gradients in specific leaf area. The increase in Ks with height corresponded to a decrease in xylem wall fraction (a measure of wall thickness), an increase in percentage of earlywood and slight increases in conduit diameter. These changes are probably adaptive responses to the increased transport requirements of leaves growing in the upper canopy where evaporative demand is greater. The lack of a safety–efficiency tradeoff may be explained by opposing height trends in the pit aperture and conduit diameter of tracheids and the major and semi-independent roles these play in determining xylem safety and efficiency, respectively.


The vasculature of woody plants is a critical link in the soil–plant–atmosphere continuum. As plants array their leaves skyward to capture light and avoid self-shading and competition, a connection must also be maintained to supplies of water and nutrients in the soil. To bring the soil- (water) and atmosphere- (CO2) derived precursors of photosynthate together in the presence of sunlight requires significant investment into transport tissues (xylem). Leaves that are best positioned to capture sunlight (i.e. those at the end of branches or in the upper crown) have the greatest demand for water and nutrients, but are borne on the longest transport paths where supply is most challenging (Ryan & Yoder 1997). This supply–demand paradox creates strong vertical and horizontal gradients in resource availability within a crown; these gradients may be particularly large in extremely tall species such as Sequoia sempervirens (D. Don), the Earth's tallest tree species.

Water transport from roots to leaves occurs under negative pressure (i.e. tension) and gradients in water availability throughout large canopies arise because water supply is resisted by two forces: friction and gravity. Vertical tension gradients due to gravity are ∼0.01 MPa m−1, equating to 1 MPa at the top of a 100 m tree. Frictional tension gradients vary according to transpiration rate, but at their peak are of similar magnitude across the whole plant as those generated by gravity. For example, the tension gradient between upper leaves and soil in large S. sempervirens trees at midday (due to friction and gravity) is approximately double that of predawn gradients (which includes only gravity, see Koch et al. 2004). Nevertheless, frictional gradients can be many times steeper than those caused by gravity over shorter distances in leaves and petioles where resistances are maximal (Sack, Streeter & Holbrook 2004) but are decreasingly steep as conduit diameters (Ewers & Zimmerman 1984) and conduit lengths (Comstock & Sperry 2000) increase with decreasing branch order toward the main axis.

If xylem tensions become too great (e.g. during drought), cavitation and subsequent embolism or blockage of xylem may result (Apfel 1971; Zimmerman 1983; Sperry & Tyree 1990). In conifers, the intertracheid pit membrane is comprised of a porous margo that surrounds a thickened torus region. Under mildly negative xylem pressure (Px) the pit membrane prevents emboli from spreading by allowing the torus to appress against the pit aperture, thus sealing off the air-filled tracheid. Air-seeding occurs when the Px is negative enough to displace the torus from its sealing position, thus allowing air to enter (Sperry & Tyree 1990; Sperry & Ikeda 1997). This air serves as a nucleus for cavitation, wherein water rapidly vaporizes, forming an embolus of air and water vapour that blocks further water transport through the tracheid.

Theoretically, cavitation poses the threat of physical damage to pit structures (Hacke, Sperry & Pittermann 2004) or conduit walls (Piquemal et al. 1998; Hacke et al. 2001a; see also Cochard et al. 2004), which can weaken resistance to further embolism (Hacke et al. 2001b). However, plants frequently subject to large xylem tensions possess xylem architecture that imparts resistance to cavitation (via altered pit membrane properties, Hacke & Sperry 2001) and to implosion (via thickening of conduit walls relative to their span, Hacke et al. 2001a, 2004; Sperry & Hacke 2004). Shortening of conduit length (Comstock & Sperry 2000) and narrowing of conduit diameter are also beneficial, not so much in terms of preventing failure, but by containing emboli and minimizing the length and cross-sectional area of the transport pathway that is interrupted.

Traits that confer ‘safety’ or increase resistance to failure during water stress may also reduce transport efficiency. The concept of a trade-off between morphological characters that confer either ‘safety’ or ‘efficiency’ is well established (Zimmerman 1983). To given some examples, short conduits may be advantageous in containing emboli, but introduce more conduit walls for water to traverse (Comstock & Sperry 2000). Conduit wall thickening resists implosion but also impinges on conductive cross-sectional area (Hacke et al. 2001a). At the pit level, increased resistance to air seeding is accomplished of a dense pit membrane, but at the cost of reduced conductivity through the pit (Hacke et al. 2004).

Broad patterns of increasing ‘safety’ and decreasing ‘efficiency’ of xylem with increasing aridity are evident (Tyree, Davis & Cochard 1994; Pockman & Sperry 2000) and there is some evidence demonstrating safety–efficiency tradeoffs within conifer species or individuals. For example, increased resistance to cavitation in Douglas-fir (Pseudostuga mensiesii Mirb.) seedlings was coupled with reduced hydraulic conductivity (Kavanagh et al. 1999). Within stems of mature Douglas-fir, the Ks of earlywood can be an order of magnitude greater than in latewood, but is more vulnerable to cavitation (Cochard 1992; Domec & Gartner 2002b; see also Domec & Gartner 2003). Among different organs of this and other conifer species, roots are typically more conductive than stems, but also show increased vulnerability to cavitation (Sperry & Ikeda 1997; Stout & Sala 2003; J. Pittermann, unpubl. data). Recently, McElrone et al. (2004) reported decreasing hydraulic conductivity (efficiency) and increasing resistance to cavitation (safety) along a gradient from deep roots to shallow roots to stems of four species, which broadly correlates with the direction of the tension gradient in the xylem (see also Pate, Jeschke & Alyward 1995). This provides some evidence of a safety–efficiency tradeoff resulting from a tension gradient within plant xylem, although comparing different plant organs that are under different developmental controls may introduce confounding influences such as mechanical functions of xylem (see Discussions by McElrone et al. 2004).

In the present study, we conjectured that the large tension gradients that exist in the xylem of very tall trees such as S. sempervirens act almost as ‘aridity gradients’ within the plant and are sufficient to generate vertical gradients in traits that dictate xylem safety and efficiency. The aim of our investigation was to measure the hydraulic characteristics of the same organ type (small branches) along height gradients in Sequoia sempervirens trees to determine if vertical tension gradients lead to increasingly ‘safe’ but decreasingly efficient (conductive) xylem with increasing height within the crown of individual trees.



Sequoia sempervirens specimens were sampled from three locations in coastal California. The first site (‘Big Basin’) was located at the southern end of the redwood distribution range (Big Basin State Park, CA, 37°10′N 122°14′W). Due to logistical constraints, a single 65 m tree on a west-facing slope (altitude 300 m) was selected for sampling at this site. We also sampled trees at two additional sites located in a south-central region of the S. sempervirens distribution range (Sonoma County, CA, 38°24′N 122°59′W). One of these sites (Sonoma-edge) was on the exposed western edge of and old-growth forest stand abutting a vineyard and here a 60 m tall specimen was selected. The third site (Sonoma-interior) was in this same stand, 200 m due east from the Sonoma-edge. Here a 67-m-tall tree was selected. At all three sites, the individual trees selected were part of a larger group of trees being monitored as part of a water relations experiment, but logistics required us to focus on one specimen at each ‘site’. Canopy access was by fixed climbing ropes. Leaf water potentials were measured using a PMS-1000 pressure chamber (PMS Instrument, Albany, Oregon) as part of a separate water relations experiment (S. S. O. Burgess & T. E. Dawson, unpublished results) and we present for reference a typical vertical gradient in leaf water potential for predawn and midday at the Sonoma site (Fig. 1). Although vertical gradients varied with time of day and season, they were invariably linear, close to the gravitational gradient at dawn (sometimes differing due to nighttime transpiration, see Burgess & Dawson 2004) and steeper at midday.

Figure 1.

A typical vertical gradient in leaf water potential at predawn and midday during May for a single Sequoia sempervirens at the Sonoma County site. Gradients varied with time of day and season, but were typically linear, close to the gravitational gradient at dawn and steeper at midday (S. S. O. Burgess and T. E. Dawson, unpublished results).

In addition to measuring vertical gradients in leaf water potential, a basic ‘within-canopy’ light profile was established for the Sonoma site by climbing two trees during cloudy conditions and using a handheld LI190 quantum light sensor (Li-Cor Instruments, Lincoln, Nebraska, USA) to make two spot measurements per height which were then averaged. Continuous measurements made in Spring (May & June) 2004 using data-logged wireless sensor arrays along eight additional crown gradients confirmed the general pattern shown in Fig. 2 (Tolle et al. 2005).

Figure 2.

A typical vertical light extinction profile derived from averaging measurements in two Sequoia sempervirens at midday, April 2002 at the Sonoma County site.

Hydraulic conductivity measurements

Five branches of similar diameter (∼5 mm) were cut at five sampling positions at specific heights within the crown of ∼60 m tall S. sempervirens specimens. Branches were chosen to be as similar as possible in age, distance from main stem (1 m), aspect and orientation. In this way we aimed to retain vertical crown position as the dominant variable and remove confounding effects of additional ‘path length’ caused by sampling further along branches beyond the main stem. Because the lowest branches in these tall specimens were generally 20–30 m from the ground, a sixth sampling position that we designated ‘understory’, was obtained by collecting branches from smaller adjacent specimens or sprouts from the parent plant. Branches were wrapped in moist paper towels, enclosed in plastic bags and transported to the laboratory for measurement. Branches were first re-cut under water and then lengths of branch (30–50 mm) that were devoid of side branches and leaves were excised. Bark was removed from the lengths of branch and replaced with wrappings of Parafilm® (Nalge Nunc International, Rochester, NY, USA) to facilitate a waterproof connection to Nalgene® (American Can Co., Greenwich, CT, USA) tubing of appropriate diameter. A 15 mmol KCl solution was prepared with filtered ‘millipore’ water and driven through the lengths of branch under pressure created by raising the reservoir of KCl solution ∼1 m above the branch. Water conducted through the branch was collected in a vial on a digital balance connected to a laptop computer to record weight changes every 30 s. Measurements were not initiated until flow rates stabilized. Conductivity was calculated from the rate of water flow (kg s−1) through a given length of stem (m) divided by the driving pressure (MPa) divided by either sapwood cross-sectional area (to yield Ks, sapwood-specific conductivity) or leaf area (to yield Kl, leaf-specific conductivity). Sapwood cross-sectional area was calculated by measuring the maximum and minimum diameter of the acropetal end of each branch segment using a digital micrometer. Pith area was subtracted from gross cross-sectional area by measuring its dimension under a dissecting microscope equipped with a stage micrometer. Leaf area was measured by harvesting all leaves attached to each branch segment and measuring projected leaf area using an LI-3100 (Li-Cor Instruments) leaf area meter. Following measurement of leaf area, leaves were dried and dry weights recorded. Leaf area and mass were used to calculate specific leaf area (SLA) and leaf area was used in conjunction with measurements of sapwood cross-sectional area to calculate the Huber value (sapwood area to leaf area ratio; Huber 1928).

Measurements of stem vulnerability to embolism

The resistance to embolism of S. sempervirens branches collected from 30 m and 50 m heights was determined from vulnerability curves, which show the relationship between xylem pressure and percentage loss of conductivity (PLC) caused by the embolism of xylem conduits. Vulnerability curves were generated using the centrifugal method which assumes that stem xylem pressure is a function of segment length and the angular velocity of the rotor (Alder et al. 1997).

S. sempervirens stems measuring ∼5 mm in diameter, were re-cut under water to a length of 142 mm, and the distal ends shaved smooth with a razor blade. The bark was left intact on the stem segments during the experiments. Prior to centrifugation, the maximum conductivity (Kmax) is typically measured which requires that a stem segment be flushed at a mild pressure to reverse any native embolism. However, conifer wood tends to exhibit reduced conductivity following flushing, so this step was eliminated (Mayr, Wolfschwenger & Bauer 2002; Pittermann & Sperry 2003; Sperry & Tyree 1990; Cochard 1992; Stout & Sala 2003). Although other techniques to reverse emboli are in use (e.g. vacuum infiltration), we are not confident that these do not suffer similar problems. Instead, the native conductivity (Knative) was treated as Kmax. Branches were collected in mid-March under well-watered conditions on a day following rain in order to maximize Knative. The assumption that Knative approached Kmax was tested using dye infiltration. Visual observation showed that the conductive wood of branch xylem was essentially fully functional prior to commencing vulnerability curve measurements. All hydraulic conductivity (K) measurements were obtained gravimetrically with a pressure head of 4–5 kPa, using distilled and filtered (0.22 µm) water.

Stems were mounted in a custom rotor designed to hold three segments, and fit a Sorvall RC-5C centrifuge (Kendro Laboratory Products, Newton, CT, USA; Alder et al. 1997). The segments were spun for three minutes at speeds that induce a known negative xylem pressure (Ψ) which was calculated from the angular velocity and length of the stem segment (Alder et al. 1997). The PLC caused by centrifugation at a known Ψ was calculated from the hydraulic conductivity measured after spinning (KΨ) relative to Knative such that:

PLC = 100 × (1 − KΨ/Knative)

The same segments were spun to progressively more negative pressures until the PLC reached 90% to 100%. Five stems at each of the 30 m and 50 m heights were used for each vulnerability curve.


Anatomical measurements were obtained from the stem material used to generate the vulnerability curves described above. Hand-cut, transverse sections were made on xylem from the centre of the stem, stained in toluidene blue, rinsed in distilled water and mounted in glycerine. The sections were viewed and photographed at 400× with a Nikon Eclipse E600 microscope and digital camera (Model RT KE, Diagnostic Instruments, Salt Lake City, UT, USA). Image analysis software (ImagePro, Media Cybernetics, Carlsbad, CA, USA) was used to measure the double wall thickness of two adjacent tracheids (t), and the tracheid lumen area, which was used to determine the equivalent circle diameter (b). These (t/b)2 data were acquired from tracheids within ± 10% of the sample's hydraulic mean tracheid diameter defined as Dc = Σdc5dc4 where dc = individual tracheid diameter (Kolb & Sperry 1999; Hacke et al. 2001a). Two to three files of tracheids were analysed within two to three outer growth rings, for a minimum of 70 tracheid measurements per stem. These data were also used to calculate the percentage of earlywood and latewood area in each stem sample, where latewood is defined as xylem conduits with a lumen radius that is less than twice the width of the double cell wall (Petty 1972). Compression wood was avoided in these measurements due to its minimal hydraulic function.

The proportion of the xylem occupied by wall material (i.e. the wall fraction) was determined relative to the total area comprised of both lumen and wall area. Ray parenchyma was excluded. These data were obtained from the digital photographs of the xylem as described above.

Pit membrane and pit aperture diameter data were acquired from radial longitudinal sections of the stem material described above. These sections were handcut, stained in toluidene blue, rinsed in distilled water, mounted in glycerine and photographed at 400×. The pit membrane and pit aperture areas were used to calculate the mean diameters of these features using ImagePro software as described above. A minimum of 170 intertracheid pits were analysed from stems at each of the 30 m and 50 m heights.


Statistical analyses were performed with Systat (version 10.0, SSI, 2000). ancova were used to test for tree/site and crown height differences in leaf-specific conductivity (Kl), sapwood-specific conductivity (Ks), Huber value, and SLA values. Values for Kl, Huber value and SLA were natural-log-transformed to meet normality assumptions. Tree/site (Sonoma Edge, Sonoma interior, Big Basin) was considered a fixed factor for all analyses. Crown height of measurement was considered a continuous covariate. All models were examined for homoscedasticity and normality of residuals.

Based on the work of Sprugel, Hinckley & Schaap (1991) and Brooks et al. (2003), we assume that local environmental conditions and plant growth (e.g. localized auxin production) strongly determine xylem hydraulics (see also Tyree 1988 for the idea that minor branches might be viewed as a collection of small independent plants each ‘rooted’ in the bole). Therefore, variables at a given crown height develop independently from other heights. As a result, crown height was not considered nested within tree/site.

Error bars on graphs purely serve to indicate repeatability of measurements using the experimental apparatus described. To gain a single data point for a given treatment (site and crown height), five similar branches from the same sampling point were averaged to reduce noise intrinsic to the measurement process. This was repeated twice, to provide two observations per sampling position.


Hydraulic conductivity and related measurements

Figure 3 shows sapwood-specific conductivity (Ks) of S. sempervirens branches for all trees/sites and all sampling occasions (2 samplings per site, n = 5 per height, per sampling). Sapwood-specific conductivity (Ks) showed a small but significant increase with height (F = 8.687, d.f. = 1, P = 0.007) in all trees sampled. There were no significant differences among trees (F = 1.378, d.f. = 2, P = 0.269) nor in the relationship between height and Ks among trees (F = 0.871, d.f. = 2, P = 0.430). However, the edge tree from the Sonoma site (see open symbols) exhibited the strongest correlation between Ks and height (R2 = 0.49).

Figure 3.

Vertical gradients in the sapwood-specific conductivity (Ks) of small (5 mm) branches in the crowns of three Sequoia sempervirens specimens measured at three sites on the north-central Californian coast. Units are expressed per m2 of xylem cross-sectional area.

Leaf-specific conductivity (Kl) also increased significantly with height (F = 52.35, d.f. = 1, P < 0.000) in all trees sampled (Fig. 4). Differences among trees/sites were significant at the P < 0.1 level (F = 2.553, d.f. = 2, P = 0.097) and there was no significant difference in the shape of the relationship between height and Kl among trees/sites (F = 1.233, d.f. = 2, P = 0.308). SLA, which has considerable bearing on measurements of Kl (Kl includes projected leaf area in the calculation and is thus affected by changes to leaf morphologies and SLA) decreased significantly and exponentially with height (F = 659.41, d.f. = 1, P < 0.000), in all trees sampled (Fig. 5). Magnitudes of SLA for any given branch height differed significantly (P < 0.05) among sites/trees (Sonoma Edge < Sonoma interior < Big Basin) (F = 12.225, d.f. = 2, P = 0.001) but the shape of the relationship between SLA and height was not significantly different among trees/sites (F = 1.606, d.f. = 2, P = 0.230).

Figure 4.

Vertical gradients in the leaf-specific conductivity (Kl) of small (5 mm) branches in the crowns of three Sequoia sempervirens specimens measured at three sites on the north-central Californian coast. Units are expressed per m2 of leaf area.

Figure 5.

Vertical gradients in SLA of leaves from the crowns of three Sequoia sempervirens specimens from three north-central Californian coastal sites.

Sapwood to leaf area ratios (Huber value) increased significantly (F = 153.92, d.f. = 1, P < 0.000) with height (Fig. 6) in all trees sampled. Differences among trees/sites were significant at the P < 0.1 level (F = 3.218, d.f. = 2, P = 0.056) but the relationship between Huber value and height did not differ significantly among trees/sites (F = 0.84, d.f. = 2, P = 0.443).

Figure 6.

Vertical gradients in sapwood area to leaf area ratios (Huber value) of branches from crowns of three Sequoia sempervirens specimens growing at three north-central Californian coastal sites.

Vulnerability to embolism

Figure 7 shows xylem vulnerability curves from branches collected at 30 m and 50 m from the Sonoma Edge tree. Branches from 50 m were more resistant to cavitation (pressure required for 50% loss of conductivity, P50 =−7.85 MPa) than branches collected at 30 m (P50 =−5.78 MPa). These values were significant at the P < 0.1 level (P = 0.0747, see Table 1). The curves diverged most obviously at xylem pressures more negative than −6 MPa, with the 50 m stems exhibiting a PLC of only 63% at −10 MPa, compared to 90% for the 30 m stems at the same pressure.

Figure 7.

Vulnerability to cavitation in small (5 mm) Sequoia sempervirens branches sampled at two heights from the crown of a single redwood tree at Sonoma County, California.

Table 1.  Summary of efficiency and safety parameters measured for branches at 30 and 50 m height in the crown of a single Sequoia sempervirens specimen growing at Sonoma County, California
 Stems at 30 mStems at 50 mP-values
  1. Data are presented as means ± SEM. *Heights in this case were 25 m, 46.6 m. †See text for statistical treatment of all branch heights.

Efficiency parameters
 Hydraulic diameter (µm)13.78 ± 0.3113.88 ± 0.18 0.82
 Earlywood area (%) 5.46 ± 10.4813.97 ± 11.02 0.41
 Mean pit membrane diameter (µm)10.82 ± 0.0810.76 ± 0.08 0.62
 Wall fraction (%) 0.58 ± 0.05 0.46 ± 0.02 0.0054
 Ks (kg m−1 MPa−1 s−1)* 0.41 ± 0.02 0.44 ± 0.05
Safety parameters
 (t/b)2 0.14 ± 0.01 0.11 ± 0.01 0.19
 Mean pit aperture diameter (µm) 3.88 ± 0.04 3.29 ± 0.03<0.0001
 Resitance to cavitation (P50, MPa)−5.78 ± 0.00−7.85 ± 0.00 0.0747
 Wood density (g/cm3) 0.69 ± 0.04 0.55 ± 0.01 0.0126


Table 1 summarizes the efficiency and safety parameters measured for branches at 30 and 50 m. The ‘hydraulic diameter’ of tracheids, percentage of xylem containing early wood and resistance to cavitation (P50) were all slightly greater in stems sampled from 50 m height compared to those sampled from 30 m (see Table 1 for statistical data). The fraction of xylem given to wall area was lower in stems from 50 m than in those from 30 m.

Wood density and tracheid (t/b)2 ratios were lower in stems from 50 m than in those from 30 m and there was a statistically significant decrease in mean pit aperture diameter, despite no change in pit membrane diameter.


Trends with height–‘efficiency’ parameters

In contrast to the predictions of our original hypothesis, we did not find a decrease in the hydraulic conductivity (‘efficiency’) of S. sempervirens branches with increasing height. Instead, using arguably the most unambiguous and mechanistic measure of hydraulic conductivity Ks, a small, but significant increase in hydraulic conductivity with height was evident. Leaf-specific conductivity (Kl), in contrast, increased exponentially with height; however, this appeared to be largely due to changes in leaf morphology. SLA decreased rapidly with height, resulting in less leaf area relative to sapwood area (increased Huber value), thus increasing K relative to leaf area (i.e. increasing Kl with height, see Figs 2, 5 and 6 and see below for further discussion). The increase in Ks with height may be best explained by the results of our microscopy work which showed a slightly greater mean hydraulic diameter of xylem conduits, a greater percentage of earlywood and reduced wall fraction in stems collected from 50 m height versus 30 m height (see Table 1).

Our finding that the hydraulic conductivity of branches increases with height in the crown is consistent with the ideas of Zimmermann (1978) who stated that the upper leaves are partly compensated for their height ‘disadvantage’ by having less overall resistance in the root–leaf hydraulic flux pathway. Very recent work by Sellin & Kupper (2005) involving experimental manipulation of water availability to upper and lower canopy shoots of Betula pendula Roth also supports the idea that upper canopy shoots are hydraulically favoured.

Our results disagree in part with some findings for other conifers. Hubbard et al. (2002) found no difference in Kl from branches sampled at 25 m and 10 m height in the crowns of ponderosa pine trees. However, owing to a reduction in sapwood area to leaf area ratio (Huber value) between these heights, we can deduce that Ks, had they reported it, would probably have increased with branch height. Other work on conifer species in the context of the hydraulic limitation hypothesis is more difficult to compare to our findings because it relates to comparisons of whole-plant Kl for different-sized trees, rather than differences between branches within individual crowns (Phillips et al. 2002). However, the general pattern of results wherein taller trees do not suffer reductions in Kl as a consequence of height and may even exhibit increased Ks (Phillips et al. 2002) broadly supports our findings that height and the effects of gravity do not inevitably lead to reductions in branch hydraulic conductivity. One explanation to come out of the above studies is that taller (and thus stouter) trees have greater capacitance, providing a more easily extractable pool of water (Domec & Gartner 2001) proximal to branches growing at height (Phillips et al. 2003). This explanation may also be relevant for S. sempervirens and is currently under investigation.

‘Safety’ parameters

The classic idea of a safety–efficiency trade-off (e.g. Zimmerman 1983) would suggest that an increase in xylem conductivity or ‘efficiency’ with height should be accompanied by reduced ‘safety’: our data do not support this idea. Although wall reinforcement (t/b)2 decreased with height, resistance to cavitation (P50) was greater in upper branches compared to lower branches. These results are in agreement with the recent findings of Koch et al. (2004; see their supplementary material) for S. sempervirens and, despite different sampling methods, they are in broad agreement with results for Douglas-fir (Domec & Gartner 2001). Although the differences between heights are small for typical minimum leaf water potentials for S. sempervirens at our site (approximately −2 MPa), they may become very important under drought conditions. Also, as Koch et al. (2004; see their supplementary material) point out, even slight increases in the risk of cavitation may be important if these authors are correct in suggesting that cavitation is very difficult to repair in these tall species.

Loss of transport capacity by wall implosion has never actually been demonstrated, whilst cavitation is a regular and well-documented danger. Thus, in our study, the data which show a decrease in wall reinforcement with height probably have less immediate relevance to xylem ‘safety’ than the data showing reduced vulnerability to cavitation with height. This increased cavitation resistance at greater heights is probably best explained by our microscopy data which showed that pit aperture decreased significantly with height. A decrease in pit aperture may increase ‘safety’ by reducing the likelihood of the torus slipping from its sealing position against the pit aperture. This may prevent air from entering the functional tracheid at very negative xylem pressures (Px), and nucleating cavitation. Hacke et al. (2004) have shown that at a broad range of Px, pit aperture scales with pit membrane diameter, thereby maintaining a generally constant overlap of the torus against the pit aperture. By contrast, our data indicate a conserved pit membrane diameter at both heights, which suggests that the torus overlap may be greater at 50 m than at 30 m. This implies that some adaptive or developmental flexibility exists in the torus overlap with regard to average or minimum native xylem pressure. Mayr, Rothart & Damon (2003) have identified similar trends in leader and nonleader shoots of Picea abies.

Safety–efficiency trade-offs

The simultaneous increase of both Ks and resistance to cavitation with height in S. sempervirens challenges a simple concept of a safety–efficiency trade-off. A number of parameters contribute to Ks and P50 values and each probably affect these values in opposite directions. However, some parameters are central and some peripheral in determining these traits. For example, a slight increase in conduit diameter will slightly and linearly increase the percentage of conductive area blocked for a given cavitation frequency, but will increase individual conduit conductivity by a fourth power function (Comstock & Sperry 2000). In other words, conduit diameter is very important to Ks but not particularly important to P50 (it doesn’t affect the mechanism of water stress-induced cavitation at all; see Tyree et al. 1994; Pittermann & Sperry 2003).

Similarly, slight increases in torus overlap may have a massive impact on the air-seeding threshold but not contribute particularly to overall tracheid resistance, especially given that pit aperture conductance scales only to the third power (in contrast to the fourth power scaling with conduit conductance, Hacke et al. 2004). Although recent modelling efforts by Hacke et al. (2004) have given some estimates of the relative contributions of parameters such as pit conductivity to overall tracheid conductivity, more work is needed to gauge the influence of changes to torus overlap on overall xylem conductivity.

Is the safety–efficiency tradeoff obscured by mechanical requirements?

Although vulnerability to cavitation was not measured in branches throughout the whole height gradient of a large redwood, from the data gathered it appears likely that branches at all heights are quite resistant to cavitation at minimum native xylem pressure (approximately −2.2 MPa as measured in the field by the pressure chamber technique). These pressures would induce < 10% loss of conductivity for branches at 30 m height, which is near to the base of live crown for many individuals. This apparent ‘overengineering’ from a hydraulic point of view may represent ‘appropriate engineering’ for other purposes such as safety against frost damage (Tyree et al. 1994) or mechanical failure (see Domec & Gartner 2002a). Unlike angiosperms where there is a clear division of labour between hydraulics (vessels) and mechanical support (fibres), conifer xylem performs both a hydraulic and a structural function in supporting the crown. For example, angiosperms generally have a safety factor from implosion of ∼2.4 whereas this safety factor is doubled in conifers to 4.8 (Hacke et al. 2001a, 2004). If mechanical requirements contribute to this large safety factor, then the decrease in (t/b)2 ratios with height might suggest reduced mechanical stresses in upper branches. Although wind exposure is no doubt greater in the upper canopy, lower branches have a greater leaf area and a more horizontal display angle (to intercept more light); such differences could increase the stress loads caused by wind and precipitation.

Koch et al. (2004) recently suggested that hydraulic overengineering is a vital feature of redwood xylem since the permanent tensions caused by gravity effects prevent repair of embolized conduits. However, existing models (Holbrook & Zwieniecki 1999; Konrad & Roth-Nebelsick 2003) and experimental data (e.g. Hack & Sperry 2003) suggest that emboli can be ‘repaired’ under mild tensions in some angiosperms; thresholds for repair in S. sempervirens have not yet been determined. Also, S. sempervirens absorbs water through its leaves during fog and rain events, resulting in downward flux of water through xylem (Burgess & Dawson 2004); this presumably relieves xylem tensions for short periods and may promote embolism repair.

The role of light in dictating water transport capacity

Crowns of S. sempervirens are exposed to strong vertical light gradients due to their high leaf area index (Westman & Whittaker 1975) and this will influence auxin production and rate of wood growth. Our results showed stems from 30 m contained only 4.44% earlywood, while 50 m stems contained 11.38% earlywood. Protz, Silins & Lieffers (2000) showed that shading in lower branches reduced the production of earlywood, average tracheid diameter and branch hydraulic conductivity. These changes were also linked to reduced stomatal conductance and death of branches in the lower crown. An apparent ‘planned senescence’ of lower branches, mediated by auxin production, makes for an interesting counterpoint to the currently popular hydraulic limitation hypothesis. Here, tree height and hydraulic path-length are considered to limit water supply to upper branches enough to reduce stomatal conductance and stop branch growth (see Koch et al. 2004 for the application of this hypothesis to S. sempervirens). The work of Zimmerman (1978) and Protz et al. (2000) suggests that upper branches in a profitable light environment are hydraulically favoured, whereas unprofitable shaded branches slowly have their water supply cut-off. This does not argue against an ultimate end-point to hydraulic supply capabilities for upper branches, only that, where biophysical constraints are not reached, light effects might produce trends opposite to those expected by the hydraulic limitations hypothesis.

Trends across sites/trees also indicate the role of light environment

Two general features were apparent among all of our measurements of Ks, Kl, Huber value, and SLA. Owing to a limited sample size, statistical differences between sites/trees were marginal, except in the case of SLA where simple measurement techniques reduced variance caused by measurement error. Despite this, a fairly consistent ranking was apparent, such that the southern ‘Big Basin’ site/tree demonstrated ‘shaded’-type traits: low hydraulic conductivity (see Feild et al. 2001), low Huber value (increased leaf area for a given amount of sapwood) and high SLA (wide, thin leaves). The Sonoma Edge site/tree exhibited ‘well-lit’-type traits (values opposite to Big Basin for the above-mentioned traits) and the Sonoma Interior site/tree exhibited intermediate values in the above traits. Our wireless microclimate monitoring (Tolle et al. 2005) and earlier anecdotal observations suggest that the light environment of these sites matches these rankings: Sonoma Edge was on the extreme western edge of a patch of old growth forest, Sonoma interior was within the old growth forest but very near the northern edge and Big Basin was in contiguous old growth growing in a steep mountainous area where a series of high ridges blocked light considerably. For all sites, the vertical trends in leaf morphology and branch hydraulic conductivity of course strongly reflect the vertical light gradient (Fig. 2) although Woodruff, Bond & Meinzer (2004) and Koch et al. (2004) remind us that light is not the only factor that drives leaf morphology – turgor pressure is important. The exponential trends in leaf-specific conductivity, SLA and Huber value appeared to more strongly reflect the exponential trend in light extinction than the linear trends in leaf water potential (Fig. 1). We are currently developing and using improved sensing techniques to measure seasonal averages of within-crown light levels (Tolle et al. 2005) that should better enable us relate hydraulic parameters to the light environment. Further experiments in fire-defoliated forests are also underway in an effort to better separate the effects of light from water potential.

In summary, it appears that the trends in hydraulic conductivity we observed can be best explained in terms of increases in earlywood (Domec & Gartner 2002a) and decreases in wall fraction with height. All of these factors are likely to be strongly linked to the light environment within the crown and apical dominance may also play a role (Zimmerman 1978). In disagreement with our original hypothesis, the standing water potential gradient alone is not sufficient to cause decreases in hydraulic efficiency with height in tall conifers, despite hydraulic safety increasing as predicted (the latter being also relevant to the larger hydrodynamic gradients in addition to the hydrostatic gradient). Further research into this apparent lack of a safety–efficiency trade-off should include other parts of the whole hydraulic pathway, not just a one-size class of branches. For example, the expected trade-off is observed within tissues of the main trunk of Ponderosa pine (Pinus ponderosa Dougl. ex Laws,Domec & Gartner 2003) and Douglas-fir (Domec & Gartner 2002b). Understanding why small branches differ in this regard may benefit from careful separation of the hydraulic drivers of xylem architecture versus the mechanical requirements associated with branch display angle, leaf area/mass, etc. On the smaller scale, continued modelling and experimental efforts to understand the relationships between pit architecture and overall conduit structure and function will prove valuable for understanding how xylem safety and efficiency covary in the crowns of large woody plants.


We thank Vanessa Schmidt, Jia Hu, and Eric Dubinsky for their laboratory and field assistance. Taylor Feild and Neil Hausman provided methodological assistance. For financial support we thank the A.W. Mellon Foundation and Global Forest (18-2000-112/113) (TD and SB). SB thanks the Australian Research council for 2004 support (DP0344310). JP acknowledges the support of the National Science and Engineering Research Council of Canada and the U.S. National Science Foundation DDIG (IBN-0308862). Comments provided by John Sperry and the Utes’ journal club improved the manuscript and we thank them. Comments by anonymous reviewers also greatly strengthened the manuscript.