Does soil nitrogen influence growth, water transport and survival of snow gum (Eucalyptus pauciflora Sieber ex Sprengel.) under CO2 enrichment?


B. J. Atwell. Fax: +61 2 98508245; e-mail:


Eucalyptus pauciflora Sieber ex Sprengel. (snow gum) was grown under ambient (370 µL L−1) and elevated (700 µL L−1) atmospheric [CO2] in open-top chambers (OTCs) in the field and temperature-controlled glasshouses. Nitrogen applications to the soil ranged from 0.1 to 2.75 g N per plant. Trees in the field at high N levels grew rapidly during summer, particularly in CO2-enriched atmosphere, but suffered high mortality during summer heatwaves. Generally, wider and more numerous secondary xylem vessels at the root–shoot junction in CO2-enriched trees conferred fourfold higher below-ground hydraulic conductance. Enhanced hydraulic capacity was typical of plants at elevated [CO2] (in which root and shoot growth was accelerated), but did not result from high N supply. However, because high rates of N application consistently made trees prone to dehydration during heatwaves, glasshouse studies were required to identify the effect of N nutrition on root development and hydraulics. While the effects of elevated [CO2] were again predominantly on hydraulic conductivity, N nutrition acted specifically by constraining deep root penetration into soil. Specifically, 15–40% shallower root systems supported marginally larger shoot canopies. Independent changes to hydraulics and root penetration have implications for survival of fertilized trees under elevated atmospheric [CO2], particularly during water stress.

elevated CO2

carbon dioxide




Little is known about how eucalypt species will respond to rising CO2 concentrations, particularly those non-commercial species that occur naturally on infertile soils. Studies on Eucalyptus tetrodonta and Eucalyptus miniata provide the only comparative growth analysis (Duff, Berryman & Eamus 1994). Eucalyptus pauciflora Sieber ex Sprengel. (snow gum) is distributed across elevated landscapes on poor soils, where it naturally grows slowly but responds to the higher N status typical of former agricultural soils (Wong, Kriedemann & Farquhar 1992). However, preliminary evidence has indicated that snow gum saplings grown in N-fertilized soils died rapidly after summer heatwaves, even when stored soil moisture was available. This observation encouraged our study of survival and growth under the atmospheric conditions that will prevail in the foreseeable future, and thus imperil stands of this important temperate woody species. Because CO2 enrichment and N fertility synergistically stimulated leaf development and productivity of mature Pinus taeda L. stands (McCarthy et al. 2006), we hypothesized that these factors could compromise hydraulic function and survival of eucalypts during southern Australian heatwaves. Moreover, it is hoped that snow gum might provide insights into survival of other non-selected eucalypt species.

The vascular system in woody plants is dynamic over periods of weeks, as xylogenesis responds to environmental conditions, but cannot respond to the sudden transpirational demand during hot, dry days. Embolism is likely to result when transpirational demand suddenly rises (Sperry & Tyree 1988), and if leaf transpiration continues, the canopy dehydrates. While long-term acclimation of the conducting system is generally considered to be highly responsive to environment (e.g. Kolb & Sperry 1999; a meta-analysis by Mencuccini 2003; Aranda, Gil & Pardos 2005), there are also reports of plant hydraulic resistance being unaffected by long periods of water stress (Cinnirella et al. 2002). Such variation in response is at least in part genetic. There are distinct quantitative differences between genotypes in characteristics of the water-conducting system (e.g. Bucci et al. 2004; Choat et al. 2005; Addington et al. 2006), and profound differences between taxa. For example, angiosperms are predicted to have conducting systems whose capacity sometimes correlates with wood density (Bucci et al. 2004), but are more sensitive to the distribution of vessel diameters (Lovisolo & Schubert 1998). In woody species such as eucalypts, vessels can be very large (>150 µm diameter) in the early wood that is laid down during rapid spring growth (Mokany et al. 2003), thereby allowing a ‘low resistance–high risk’ pathway for sap flow to leaves (Hacke & Sperry 2001). This scaling relationship between conductance and leaf area was outlined by Mencuccini (2003) and has its origins in the ‘unit pipe model’ of Shinozaki et al. (1964). However, strong seasonal responses in vessel dimensions suggest that eucalypts might have substantial capacity to modify hydraulic architecture under suboptimal growing conditions.

Many data on plant hydraulics are derived from measurements of individual organs such as petioles (Bucci et al. 2003) and stems (Lovisolo & Schubert 1998; Medhurst & Beadle 2003). Fewer papers report whole-plant hydraulics calculated using the Ohm's law analogue (e.g. Eamus, Berryman & Duff 1995; Brodribb & Hill 2000; Aranda et al. 2005; Lovelock et al. 2006), and the contribution of roots to hydraulics in woody plants comes largely from recent studies on perennials from largely xeric sites (e.g. Martínez-Vilalta et al. 2002, 2007; McElrone et al. 2004; Addington et al. 2006). While the hydraulic resistance imposed by the xylem vasculature in shoots and stomatal responses in leaves are important, roots also impose substantial resistances to sap flow in transpiring plants through the radial and axial pathways of water flow (e.g. Landsberg, Blanchard & Warrit 1976; Roberts 1977; Running 1980; Nardini & Pitt 1999; Brodribb & Hill 2000; Martínez-Vilalta et al. 2007). Addington et al. (2006) showed specifically that the hydraulic adjustments in Pinus palustris from xeric sites must partly result from hydraulic changes in roots. Roots are also subject to embolism and can thus incur hydraulic failure (Tsuda & Tyree 1997; Froux et al. 2005; Domec, Lachenbruch & Meinzer 2006) leading to plant death.

The hydraulic conductance of herbaceous plants does not respond uniformly to elevated [CO2] (Bunce & Ziska 1998). In potted woody plants, Heath, Kerstiens & Tyree (1997) showed that CO2 enrichment reduced leaf area-specific hydraulic conductance of stems of oak, but had no effect on beech. Eamus et al. (1995) also reported a reduction in ‘whole-shoot conductivity’ (calculated from the relationship between transpiration and water potential) when 20-month-old woody perennials were grown at elevated [CO2]. Reports on individual organs such as branches show that in oak trees surrounding natural CO2 vents in Italy, there were no effects of elevated [CO2] on specific conductivity of branches (Tognetti, Longobucco & Raschi 1999). Atkinson & Taylor (1996) measured species differences in stem conductivity, with segments of oak stems being able to conduct more sap when grown at elevated [CO2], while cherry stems were unaffected by [CO2]. Interestingly, xylem-specific conductivity (XSC) (transport corrected for xylem area) was independent of [CO2] treatment in Atkinson and Taylor's study, supporting the view that mass flow through the xylem elements was achieved with the same efficacy when stems grew under CO2 enrichment in controlled glasshouse conditions. Because elevated [CO2] improves water-use efficiency at the leaf level, it might be expected to alter whole-plant hydraulic characteristics (Bunce & Ziska 1998; Sperry et al. 2002; Wullschleger, Tschaplinski & Norby 2002). However, in practice, measured hydraulic effects are inconsistent among data sets from shoots, and data do not extend to root systems.

Root system architecture responds to soil water deficits and nutrient deprivation (Ho et al. 2005), setting up a spatial conflict between exploration for nutrients near the soil surface and water deeper within the profile. Effects of individual abiotic factors (e.g. drought, nutrient status) on the key determinants of water transport vary widely, explaining why we cannot easily predict how field conditions will affect water acquisition. Drought constrains water use and, accordingly, the capacity of the conducting pathways is reduced through narrowing of vessels (Lovisolo & Schubert 1998; Corcuera, Camarero & Gil-Pelegrin 2004). However, this is still not widely tested across species. The effects of inorganic nutrition on hydraulics are even more equivocal. While N fertilization slightly increased mean xylem vessel diameter and susceptibility to embolism in poplar, P supplementation exerted a protective effect on hydraulics through pit membrane development (Harvey & van den Driessche 1999) and enhanced hydraulic conductivity in mangrove (Lovelock et al. 2006). Addition of inorganic fertilizer without irrigation reduced leaf-specific conductivity (LSC) and risk of embolism in the conifer P. taeda (Ewers, Oren & Sperry 2000), while high N supply improved water-use efficiency in Eucalyptus grandis without affecting plant hydraulics (Clearwater & Meinzer 2001).

Elevated [CO2] also stimulates fine root growth, and thus, root distribution, although the effect might be peculiar to water- and nutrient-limited soils (Wullschleger et al. 2002) and is confounded by ontogeny (Norby et al. 1999). Elevated [CO2] also caused a modest increase in vessel number and diameter in stems of oak, a ring-porous species but not cherry, a diffuse-porous species (Atkinson & Taylor 1996), encouraging us to investigate CO2 enrichment effects in E. pauciflora, with its distinct ring-porous wood. Nothing is known of the coordination of architecture and hydraulics in woody root systems. In the experiments reported here, we attempt to distinguish between specific effects induced by elevated CO2 from accelerated development (ontogeny) when analyzing the impact of N nutrition and elevated [CO2] on water transport.

The present study combined measurements of snow gum grown under ambient and elevated [CO2], either in open-top chambers (OTCs) subject to the natural vagaries of the weather or in controlled glasshouse conditions. We sought to examine the contrast in survival of trees when fertilized with N in CO2-enriched atmospheres by investigating the growth and hydraulics of roots and stems. Water transport was assessed directly in root systems by applying mild suctions to intact root systems immediately after soil saturation (see Brodribb & Hill 2000), and indirectly by measuring radial xylem dimensions and applying the theory of the Hagen–Poiseuille equation.


Growth conditions – field

Seeds of E. pauciflora Sieber ex Sprengel. were collected from trees in the Gudgenby Valley, Australian Capital Territory (ACT), elevation: 1000 m. The seeds were cold-stratified under moist conditions at 4 °C for 4 weeks before germinating on potting mix in a glasshouse. The seedlings were grown for 2 weeks and then transferred to seedling tubes which remained in the glasshouse for a further 6 weeks. They were then moved outside and grown under natural conditions in Canberra, ACT (elevation: 700 m) for 2 months. A selection of plants with uniform size and leaf number was transplanted directly into PVC cylinders 400 mm deep and 150 mm in diameter which were buried with the top flush with the soil surface within OTCs at a field site near Bungendore, New South Wales (35.25°S, 149.43°E). The soil at the site was a brown orthic Tenosol (Uc5.31) (light sandy clay loam) with 150 mm of grey sandy upper horizon that was visibly darkened by the presence of organic matter. To mimic the original soil profile, the top 150 mm of each cyclinder was filled with topsoil while the rest of the cylinder was filled with subsoil. Basal fertilizer was mixed through the topsoil: it comprised a basal application of 32 g superphosphate; 123 g potash; 61 g magnesium sulphate plus three rates of N: low N (0.10 g N per plant), intermediate N (0.75 g N per plant) and high N (2.75 g N per plant) in the form of ‘slow release’ Osmocote 23:0:0 (containing ammonium nitrate and calcium carbonate). The soil surface was covered with straw throughout the experiment. The OTCs were subject to either ambient CO2 conditions (370 µL L−1) or flushed with CO2 to achieve a mean concentration of 700 ± 50 µL L−1 in the atmosphere during each day.

Ten OTCs were installed with each of the ambient and elevated [CO2] treatments represented by five replicates. Chambers were essentially as described by Ashenden, Baxter & Rafarel (1992) and as modified by Lutze et al. (1998): each chamber was 1.28 m in diameter and 1 m high. The [CO2] in the chambers was monitored frequently throughout day-light hours with an IRGA (Li 6400, Li-Cor, Lincoln, NE, USA). One hundred and twenty seedlings of equal height and leaf area were selected to be planted into OTCs, with 12 seedlings randomly allocated to each chamber.

In each OTC, N was added at the rates described earlier to each of four separate cylinders. Three plants, one from each N level, were sacrificed from each OTC in midwinter, with further destructive harvests in November (spring) and February (summer). The summer harvest involved twice as many plants as each earlier harvest (two plants sacrificed per N level from each OTC). The destructive hydraulic measurements coincided with the harvests.

Experiments were run in successive seasons on adjacent locations. The seedlings were transplanted into damp soil on 7 April 1999 and 28 March 2000, but did not receive additional water at any time during plant growth up until harvest a year later. Trees grew faster in 2000–2001, which was a wetter season, but the effects of [CO2] and N were very similar in both seasons. Survival and vascular characteristics are reported for trees from both seasons (e.g. Table 1).

Table 1.  Vessel dimensions and associated calculations in field- and glasshouse-grown snow gums at the highest and lowest soil levels of N, and ambient and CO2-enriched atmospheres
YearSeasonTreatmentnVessels per (root or stem) axisMean lumen area (µm2)inline imageper transverse section × 10−6 (µm4)Vessel frequency (vessels mm−2)% of sapwood area occupied by lumina
  1. None of the root characteristics showed a CO2 × N interaction, so one-way analyses of variance (anovas) are reported. Vessel dimensions for both harvests in the glasshouse were combined when presenting effects of N levels, because there was no significant interaction between N levels and harvest on vessel dimensions. Values are means ± standard error. One-way anova was used to establish significance of CO2 and N treatment comparisons at each harvest.

  2. n.s., not significant.

Root–shoot junction – field
Year 1Late summerAmbient5262 ± 62315 ± 188334 ± 4029 ± 86.7 ± 1.1
Elevated6291 ± 652710 ± 90457 ± 6519 ± 75.1 ± 0.7
  n.s.F = 4.85; P = 0.053F = 11.4; P = 0.015n.s.n.s.
Low N4240 ± 312588 ± 183205 ± 3619 ± 104.9 ± 2.9
High N5329 ± 803000 ± 210371 ± 7822 ± 96.6 ± 3.2
  n.s.F = 3.67; P = 0.091F = 3.91; P = 0.088n.s.n.s.
Year 2Late springAmbient10498 ± 55736 ± 4243 ± 1043 ± 44.3 ± 0.2
Elevated10464 ± 75856 ± 6652 ± 1438 ± 64.0 ± 0.3
Low N9414 ± 44838 ± 6238 ± 635 ± 42.9 ± 0.5
High N10583 ± 82815 ± 6157 ± 1746 ± 63.7 ± 0.8
Year 2Late summerAmbient51102 ± 183944 ± 135180 ± 7745 ± 63.4 ± 0.6
Elevated161732 ± 2231207 ± 92564 ± 16734 ± 33.2 ± 0.3
  n.s.n.s.F = 3.27; P = 0.095n.s.n.s.
Low N91592 ± 1371056 ± 266280 ± 12239 ± 44.1 ± 1.5
High N111618 ± 1731577 ± 315445 ± 10934 ± 55.4 ± 1.9
Stems – field
Year 1Late summerAmbient6331 ± 58973 ± 731.8 ± 2.463 ± 14.5 ± 0.07
Elevated13331 ± 271005 ± 1232.3 ± 2.650 ± 34.7 ± 0.07
Year 2Late springAmbient10387 ± 31718 ± 4030.2 ± 5.452 ± 24.2 ± 0.3
Elevated10330 ± 41735 ± 5528.2 ± 6.450 ± 54.3 ± 0.4
Year 2Late summerAmbient5348 ± 43750 ± 11232.5 ± 13.169 ± 105.4 ± 0.3
Elevated16297 ± 22719 ± 6225.4 ± 5.458 ± 54.4 ± 0.3
Root–shoot junction – glasshouse
Early summerAmbient8333 ± 51208 ± 2149 ± 487 ± 810.5 ± 0.9
Elevated8483 ± 601264 ± 1378 ± 1966 ± 98.7 ± 1.4
  F = 8.06; P = 0.012n.s.F = 2.98; P = 0.106n.s.n.s.
Late summerAmbient8650 ± 601283 ± 1108 ± 1268 ± 118.4 ± 1.1
Elevated8712 ± 881125 ± 1100 ± 854 ± 106.0 ± 1.1
N in combined harvestsLow N16435 ± 421277 ± 7477 ± 161 ± 97.8 ± 1.2
High N16542 ± 661191 ± 5780 ± 1177 ± 79.2 ± 1.3

Because field-grown snow gums are relatively variable, and recovery of intact root systems was impossible, the next phase of experiments was aimed at studying glasshouse-grown plants in large columns replicating the field soil structure. Temperature and atmospheric [CO2] could be controlled, hydraulic conductance was measured more conveniently and roots recovered. The purpose was to determine whether root system conductance and root penetration might have made plants at high N and/or elevated [CO2] more vulnerable to dehydration during summer.

Growth conditions – glasshouse

Plants were grown in large PVC cylinders (1200 mm tall × 220 mm diameter), which had been split longitudinally into half cylinders then retaped to form entire cylinders. The inner walls were coated with strong adhesive to which soil was glued in order to prevent roots growing preferentially down the sides of pots. Topsoil (0–150 mm) and subsoil were excavated separately from the Bungendore field site, and transported to Sydney where the topsoil was mixed thoroughly with granular fertilizer at the ‘high’ and ‘low’ rates identical to the field experiment (the intermediate N level was not applied.) Pots were then repacked to mimic the proportions of the field soil. The glasshouse experiment was performed with paired glasshouses in which treatments (and therefore, the CO2 source) and plants were swapped weekly to avoid site effects as described in Atwell et al. (2007). Sixteen pots were placed in each of two glasshouses, with one set of 16 pots maintained at ambient atmospheric conditions and the other set in atmospheres enriched to 700 µL L−1 by continuous control through a solenoid system with a ±50 µL L−1 deviation. Pairs of replicates (one ‘high’ and one ‘low’ N treatment) were randomized within glasshouses.

Glasshouses were air conditioned at temperatures set at 24–10 °C (12 h day–night cycle) and vigorously ventilated with large fans. Temperature deviations were monitored and never exceeded 2 °C and rarely exceeded 1 °C; in other words, the two glasshouses were well matched for light and temperature. Photosynthetically active radiation incident on a horizontal sensor averaged 800–1000 µmol m−2 s−1 in winter, and reached peaks of 1600–1800 µmol m−2 s−1 in summer (November to February) when harvests were made. The plants were watered every second day to field capacity [24% (w/v)] by adding a quantity of water that was estimated from three spare pots which could be loaded onto a 200 kg floor balance. Oaten straw was also applied to the soil surface as we did in the field. In this way, surface soil never dried out, and yet, there was no through drainage of water from the soil column.

Seeds of E. pauciflora from the same provenance as those grown in the field experiments were sown into the base of small lengths of silicone tubing (35 mm long; 7 mm i.d.) which were propped vertically on top of the soil. The lower 5–10 mm of these tubes contained fine-mesh vermiculite to maintain moisture and ensure good germination. Three seedlings were established in each pot over the following weeks, with their stem bases ensheathed by the tube and their leaves aloft. When established, two seedlings were sacrificed and one was grown to the end of the experiment when the tube sheathing the stem gave us the opportunity of sealing the stems reliably into the hydraulic system (details as follows).

Hydraulic measurements and harvests took place 4 and 6 months after sowing, in mid- and late summer. By the time of the first harvest, the small trees were branched, semi-prostrate and had stems about 30 cm long.

Hydraulic measurements

Root systems in both field and glasshouse treatments were flooded for 12 h prior to hydraulic measurements. In the glasshouse, this entailed shrouding pots with heavy-duty plastic bags. At the cool night temperatures and low-soil organic matter levels in the flooded soil, this period was insufficient to induce anaerobic conditions, but was sufficient to eliminate major air pockets in the soil and to reduce the chance of embolism by raising soil water potential close to zero. We then investigated the hydraulics of the entire root system by removing the stems for growth analysis and connecting the remaining root stump to a controlled vacuum.

In the field, the root–shoot junction was cut with secateurs under clean water that was contained in a split funnel to flood the cut site. The cut surface was then shaved with a new razor blade, and a tubing of appropriate dimensions was chosen and fitted tightly over the cut stump. A liberal application of silicone grease was applied to the bottom end of the tubing to reduce the chance of leaks. In the glasshouse, the silicone tubing through which trees had grown now fitted snugly around the stem: it was rolled back and the stem was cut halfway down the tube to ensure a perfect seal with the ‘flow apparatus’ described later. In all cases, the cut surface was covered throughout the procedure with 5–10 mm water. A series of reducers was used to attach the tubing exiting the cut stump to a smaller-diameter tubing that had 2 mm rigid walls and internal diameters of 5–7 mm. This finer tubing connected in turn to: (1) a ‘T’ junction to which a vacuum gauge on a blind side line was connected; (2) a 0.1 mL glass micropipette that was cut to include only the graduated length (1 µL divisions); (3) a Büchner funnel to capture soapy solution exiting the pipette; (4) a side line open to the atmosphere with a tubing of different diameters in order to bleed air into the line and modulate the vacuum up to –93 kPa (±2 kPa); and finally (5) a two-stage vacuum pump. At the base of the micropipette (2, above), a rubber Pasteur pipette teat was located just below the air stream and filled with 1 mL detergent solution in order to generate bubbles that were carried in the gas flow past the graduations of the micropipette. This allowed the flow rate to be calculated by timing the passage of menisci with a stopwatch.

Root conductance was measured by simulating the natural direction of sap flow with mild suctions applied to flooded root systems (analogous to a suction method for shoots used by Macinnis-Ng, McClenahan & Eamus (2004) and similar to the method employed for roots by Brodribb & Hill (2000). Failed seals at the plant stem in the field and glasshouse that resulted in conspicuous leaks and huge flow rates were easily eliminated. Because 2–3 mL of liquid perched on top of the cut stump was visible through the tubing, suctions could be kept low enough to minimize the appearance of bubbles in the stream. Soap bubbles were used to measure flow over short intervals; the micropipettes allowed delivery of as little as 5 µL sap to be registered. Suctions were increased sequentially to −60 to −80 kPa, and then dropped back to low levels for a final measurement to confirm that resistance had not changed over time. Flow rates were constant at each level of suction over the total period of measurement (5–7 min). This approach contrasts with the alternative high-pressure flowmeter method (Sperry, Donnelly & Tyree 1988). Only flows achieved at suctions exceeding −60 kPa are reported as lower flow rates were generally unreliable. Root segments were then analysed microscopically.

Stem conductance was measured on material collected from the field and placed immediately on ice and kept wet. Branches were removed from the ice, ends trimmed and unbranched segments 15–25 cm long were placed in a pressure bomb with the cut stump in a reservoir of filtered 10 mm KCl that had been degassed. The stems were subjected to three high pressure flushes (0.15 MPa) for 20 min, then subjected to successively increasing pressures in 0.05 MPa increments. Sap flow was measured by timing the rise of sap in a 5 mL graduated pipette attached to the top of the section with tubing. The meniscus was never allowed to rise more than 10 cm above the cut stump.


Segments were then sampled for microscopy. A 2–3 cm section taken 10 cm above the root–stem junction was removed and kept in 30% ethanol until it was sectioned for xylem dimensions. At the time of sectioning, wood was washed, placed in a sliding microtome and cut transversely to produce 100 µm sections, stained briefly with 1% (w/v) toluidine blue and photographed digitally under a light microscope. The grey-scale images were then analysed using NIH Object-Image V 2.01 software with macros for vessel analysis that we have developed (Atwell et al. 2007). Our approach generated large amounts of data with 200–500 vessels per image. Multiple images were made into a single montage. For each montage image, we selected an image threshold which most accurately selected the majority of secondary xylem vessels (whose identity we independently determined under the light microscope). This threshold was then used to measure xylem lumen numbers and areas automatically across each stem section. We used the histogram of vessel areas generated from these data from each section to calculate the sum of the radius to the fourth power (inline image) by initially deriving a radius (using the formula: inline imagewhere A is the xylem lumen area) for each vessel, converting this to micrometres, raising this to the power of four and summing for all vessels in the stem section to reflect the potential flow characteristics of whole stems. The variable inline imageis proportional to the flow rate at a fixed suction gradient and originates from vessel cross-sectional areas (proportional to r2) and conductance (proportional to r2 in the Hagen–Poiseuille equation). It should be emphasized that this theoretical analysis is appraised against the flow rates induced in vivo by the suction experiments.


Growth data from the glasshouse experiment (Table 2) were analysed using three-way analyses of variance (anovas) with [CO2], soil N and harvest as main effects. Data were transformed for some variables prior to analyses to normalize distributions. There were four replicates of each measurement in the glasshouse experiment. Numbers of replicates and individuals within treatment groups in the field experiment varied because of mortality. Two-way anovas revealed no interaction between [CO2] and N supply, so the vascular data in Table 1 were analysed by one-way anovas to examine the main effects of [CO2] and N supply. Correlations and Pearson coefficients are reported for regressions of sap flow data. Standard deviations are used to represent treatment differences in Fig. 1. Differences are described as significant in the text when P ≤ 0.05.

Table 2.  Growth analysis of glasshouse-grown trees with controlled temperature and CO2 atmospheres: trees were sampled in December (early summer) and February (mid–late summer)
CO2N (g per pot)Branches (per plant)Leaves (per plant)Leaf area (cm2)Shoot fresh biomass (g)Shoot dry biomass (g)Root dry biomass (g)Maximum root depth (cm)Shoot biomass :  root biomassShoot biomass : root length (g cm−1)
  1. Trees were exposed to ambient or elevated CO2 from germination, and grown in soil that was adjusted to replicate the highest and lowest N application rates that were used in the field. Data presented are treatment means with standard errors. Three-way analysis of variance (anova) was used to establish significance effects of CO2, N and harvest treatment.

  2. n.s., not significant.

Harvest 1
AmbientLow0.6 (0.3)17 (4)281 (48)20.7 (3.1)5.3 (1.0)1.54 (0.21)97 (16)3.44 (0.28)0.06 (0.01)
High1.0 (0.5)18 (3)385 (62)25.6 (5.3)7.8 (1.1)1.49 (0.35)84 (8)5.23 (1.13)0.09 (0.01)
ElevatedLow1.3 (0.2)23 (8)491 (75)30.9 (5.3)8.8 (1.6)2.04 (0.39)120 (5)4.31 (0.42)0.07 (0.01)
High2.2 (0.9)31 (5)589 (66)39.2 (4.8)12.2 (1.9)3.03 (0.88)62 (13)4.03 (0.90)0.20 (0.05)
Harvest 2
AmbientLow2.0 (0.9)54 (4)754 (120)54.5 (6.8)15.3 (2.0)4.26 (0.72)128 (5)3.59 (0.22)0.12 (0.02)
High2.6 (0.6)62 (3)897 (130)62.5 (11.0)19.5 (3.8)3.01 (0.81)105 (9)6.48 (0.72)0.19 (0.03)
ElevatedLow1.9 (0.8)70 (12)1139 (176)81.1 (11.4)24.2 (3.2)6.73 (1.3)137 (9)3.59 (0.51)0.18 (0.02)
High3.3 (0.5)66 (8)1181 (259)76.1 (16.5)30.8 (4.8)7.61 (1.6)94 (14)4.05 (0.90)0.33 (0.08)
Significant effects (P < 0.1)
CO2 × Nn.s.n.s.n.s.n.s.n.s.n.s.0.0310.055n.s.
CO2 × harvestn.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.
N × harvestn.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.
CO2 × N × harvestn.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.n.s.
Figure 1.

Time-dependent change in the probability of seedling survival from early spring to late summer. Two upper panels show daily maximum temperatures in open-top chambers (OTCs) in the two field seasons. Circles denote the first occasion on which temperatures rose above 35 °C. Survival (lower panel) for trees averaged across these 2 years and CO2 treatments (700 µL L−1 or ambient CO2 conditions). Trees were fertilized with three rates of commercial nitrogenous fertilizer added to the soil cylinders at rates of 0.1 g N per plant (low N, open circles); 0.75 g N per plant (medium N, half-filled circles); and 2.75 g N per plant (high N, filled circles). The 95% confidence limits are shown as shaded areas.


Field responses

Stems of trees grown at elevated [CO2] and high soil N elongated rapidly during late summer, even though only 10% of those fertilized heavily with N survived until mid-February (Fig. 1). This bolting growth and mortality at high N suggest a ‘high return for high risk’ strategy for these trees, and raise questions about the root–shoot functional equilibrium and hydraulics of these plants.

In the field, N fertilization and elevated [CO2] generally increased the mean radius of xylem vessels at the root–shoot junction, in turn increasing inline imageand thus potential sap flow along the axial pathway. Specifically, these treatments caused wider xylem vessels to form by late summer, while they had little impact on vessel dimensions in early summer (Table 1). For example, late in the summer of 2000–2001 (year 2), mean vessel cross-sectional areas at the root–shoot junction were increased by about 25% through CO2 enrichment, and about 50% through N fertilization. While there were small differences in vessel number in trees from different N and CO2 treatments, they were modest except in late summer, when the stems of trees under elevated [CO2] thickened rapidly and thus had many more vessels. On the other hand, the density of vessels (vessels per unit sapwood area) and the proportion of stem area occupied by vessels were compared across treatments at each sampling time (spring and summer), but were never significantly different. Therefore, individual vessel areas, and vessel numbers and their potential effects on flow (via inline image) were considered the principal variables. They were compared in Table 1 with N level and [CO2] as independent variables because there was no significant interaction between the effects of CO2 and N.

Stems of the trees described in Table 1 were selected for a similar analysis, each being a single stem selected from multiple branches. Vessel dimensions in these individual stems were unaffected by treatment, in strong contrast to those at the root–shoot junction, suggesting that the hydraulic capacity of trees was plastic at the root–shoot junction but less variable in individual aerial branches. The increased branching of stems (see glasshouse experiment) in N-fertilized plants at elevated [CO2] was a more important factor in controlling shoot water transport than the hydraulics of individual branches (Fig. 1).

The sapwood-specific hydraulic conductance of trees grown under elevated [CO2] was almost four times greater than in trees from ambient conditions by late in the summer of 2001 (see slopes in Fig. 2). This suggests that the root systems of trees that had grown rapidly under elevated [CO2] had a commensurately greater conducting capacity in order to supply larger leaf canopies.

Figure 2.

Sap flow rates in root systems under applied suction. Sap flow rate per unit of sapwood area versus suction applied to the root–shoot junction of trees grown under ambient (inline image) and CO2-enriched (□) atmospheres in the field. The slopes of these regressions equate to sapwood-specific hydraulic conductance (µL mm−2 kPa−1 min−1). The data presented are pooled observations from trees growing at high and low N, because we determined that there was no effect of N fertilization on the hydraulics of trees. Standard errors of the sapwood-specific hydraulic conductance (µL mm−2 kPa−1 min−1) are 0.006 (ambient) and 0.013 (elevated CO2). Regression analysis using 99% confidence limits gave the range of conductance (µL mm−2 kPa−1 min−1) as 0.038–0.072 (ambient) and 0.163–0.230 (elevated CO2).

To establish whether xylem vessel dimensions were responsible for enhanced root system conductance, mean cross-sectional vessel areas at the root–shoot junction were plotted against xylem-specific conductance of whole root systems (Fig. 3). Notwithstanding quite high variation between individual plants, there was a tight linear relationship between vessel area at the root–shoot junction and XSC, indicating that larger vessels supported greater potential flow through roots. Most of the variation arose through year-to-year changes in vessel sizes.

Figure 3.

Conductivity of root systems and stem segments across a range of mean vessel areas. Natural logarithm of xylem-specific conductance (µL mm−2 lumen area kPa−1 s−1) versus mean lumen area (µm2 per vessel) of the root–shoot junction (squares) and stem segments (diamonds) of trees grown under ambient (filled symbols) and CO2-enriched (open symbols) atmospheres in open-top chambers (OTCs). Data are presented from trees sampled in two seasons (year 1, 1999–2000 season; year 2, 2000–2001 season). The regression is based only on data for the root–shoot junction (squares).

Stems of both field-grown and glasshouse-grown snow gums were highly branched, and thus hydraulics of individual stem segments had to be measured, contrasting with measurements at the root–shoot junction that integrate hydraulics of intact root systems. The relationship for stem branches (Fig. 3) illustrates that there is hardly any variance in mean vessel area and only a twofold variance in XSC, again reflecting the strong influence of branch numbers on conductance compared to conductance of individual branches.

Glasshouse responses

The growth analyses for glasshouse-grown plants are presented in Table 2. Shoot biomass was about 60% greater when the atmosphere was CO2 enriched across all N treatments. That is, the CO2 fertilization effect commonly seen in young C3 plants was evident. The effect was greater in roots, which were almost 90% heavier when trees had been CO2 enriched (P ≤ 0.01). While higher nitrogen supply generally resulted in heavier above-ground biomass, neither N alone nor the N × elevated [CO2] interaction was significant, indicating that there was sufficient N available in both N regimes to sustain the CO2 fertilization effect.

However, the most notable effect of N was that roots penetrated much deeper in soil at low N status, regardless of [CO2] supply. Plants grown at high soil N levels had especially shallow roots when they had grown in CO2-enriched atmospheres. Hence, roots of these plants would have been ineffective at extracting nitrate and water that had leached deep into the soil column. Notwithstanding the effect on length, root biomass did not show any effect of N, suggesting that shallow root systems branched or thickened more strongly in shallow soil layers. Consequently, as demonstrated by shoot biomass and shoot : root ratios, plants with high N supply allocated significantly more to above-ground biomass than those grown under low N conditions. The significant interaction between N and CO2 treatments indicates this effect was particularly evident for plants grown under high N and ambient [CO2] (Table 2). In contrast, shoot biomass : root length ratios were independently affected by elevated CO2 and high N: elevated CO2 acted by stimulating shoot growth and N by diminishing root depth.

Effects on the depth of root penetration raised the question of whether the conducting properties of root systems might contribute further to sudden and catastrophic water deficits. Hydraulics of the root system was determined by the same method as for field-grown trees, and should be considered in the light of the contrast in xylem dimensions reported in Table 1.

While N fertilization affected shoot and root growth strongly when trees grew in OTCs in the field, the effects were less marked in the glasshouse, where temperature did not exceed 24 °C and the rapid growth typical of late summer in the field was suppressed (Table 2). Table 1 shows that there were only modest effects of N supply on vessel dimensions at the root–shoot junction in trees from the glasshouse and in early summer in the field; in these trees, N therefore had no effect on potential sap flow (see inline imagein Table 1). Furthermore, actual sap flux induced by mild suction at the root–shoot junction in glasshouse-grown trees suggested a small but insignificant increase in conductance in trees grown at high N status, consistent with the absence of any effect of N supply on mean xylem vessel areas. By contrast, the significant increase in potential sap flow in field-grown trees late in the summer (reflected in Fig. 3 and the inline imagevalues in Table 1) is a clear manifestation of a surge of growth coinciding with more and larger xylem vessels, and hence, increased hydraulic demand.

Sapwood-specific conductivity (SSC) was not increased by elevated [CO2], consistent with the absence of large secondary xylem vessels when trees were grown at elevated [CO2] in glasshouses, even in late summer (Fig. 4). This might be considered surprising alongside the significant stimulus to leaf area under elevated [CO2] (Table 1), because a commensurate increase in transpiration might have been expected in larger canopies (thus greater xylem conductivity). However, leaf-level water conductance measured with gas exchange techniques showed 25–30% decrease in water loss per unit of leaf area in CO2-enriched plants (data not reported). Thus, the additional transpiring leaf area caused by enhanced leaf expansion under elevated [CO2] was largely offset by higher water-use efficiency.

Figure 4.

Specific hydraulic conductivities (µL mm−1 kPa−1 s−1) calculated from regressions of sap flow rate (µL s−1) × taproot length (mm) per unit area (mm2) against suction (kPa). The areas used were: (1) sapwood area [cross-sectional area of stems, less bark – sapwood-specific conductivity (SSC)], top row; (2) leaf area [leaf-specific conductivity (LSC)], middle row; and (3) total xylem lumen area [xylem-specific conductivity (XSC)], bottom row. Data are presented from trees that grew either in ambient (left column) or elevated CO2 (right column) conditions. Data points from plants at high and low N were pooled because N had no effect on conductivities. Regression equations are quoted, with the slope equating to the specific hydraulic conductivity (marked by asterisks). Standard errors of the SSC are 0.21 (ambient) and 0.10 (elevated CO2). Regression analysis using 99% confidence limits gave the range of conductance (µL mm−1  kPa−1 s−1) as 1.02–2.20 (ambient) and 1.01–1.81 (elevated CO2); hence, SSC not significantly different. Standard errors of the LSC are 0.0004 (ambient) and 0.0006 (elevated CO2). Regression analysis using 99% confidence limits gave the range of conductance (µL mm−1 kPa−1 s−1) as 0.0009–0.0031 (ambient) and 0.0015–0.0045 (elevated CO2); hence, LSC not significantly different. Standard errors of the XSC are 6 (ambient) and 19 (elevated CO2). Regression analysis using 99% confidence limits gave the range of conductance (µL mm−1 kPa−1 s−1) as 91–123 (ambient) and 170–272 (elevated CO2); hence, XSC differed significantly between CO2 treatments.

LSC and XSC were estimated by correcting for the respective areas of leaves and xylem lumina (instead of total sapwood area), and thus expressing water transport on a functional basis. Leaves and roots of some individuals could not be sampled, leading to variation in estimates of suction required to achieve flow (SSC versus LSC and XSC). However, we do not believe that this affected slopes, and thus, conductivity estimates. While LSC did not change in response to elevated [CO2], XSC was significantly greater in the root systems of trees under elevated [CO2]. Increases in XSC were also recorded in root systems of trees at elevated CO2 when grown in the field (Fig. 3, note log scale). Thus, those xylem vessels that specifically developed under elevated [CO2] had a greater inherent capacity to deliver sap to shoots.


In the present study, high shoot and root growth rates were induced in snow gum by elevated atmospheric [CO2] as previously shown by Wong et al. (1992) with snow gum in pots. As predicted by Norby (1994) and Tissue, Thomas & Strain (1997), elevated [CO2] did not disrupt the biomass distribution between roots and shoots, whereas low N reduced shoot : root ratios, as observed in classical experiments on other species (Brouwer 1963, 1983). Elevated [CO2] caused root systems to be heavier but shallower under high N supply, demonstrating that increased below-ground biomass allocation at high [CO2] did not compensate for the inhibitory effects of N fertilization on development of root systems. Carbon dioxide enrichment is known to accelerate shoot development, often causing denser wood to form (Norby et al. 1999), but it was not the key factor in root development in snow gum. In species that are adapted to naturally infertile soils, there is a high risk associated with the radical changes in biomass allocation as atmospheric [CO2] rises, particularly when this allocation is disrupted by fertilizers under agro-forestry. Sudden water deficits are a hazard for woody plants that must survive the dry, hot summers that characterize southern Australia. Our study of snow gum shows that root systems maladapted for efficient water extraction could be the paradoxical outcome of rapid shoot development, and thus predispose these plants to increased risk of catastrophic water deficit.

Our field experiments conducted in OTCs illustrated the selective death of trees at high [CO2] and N levels when day-time maximum temperatures exceeded 35 °C during the dry summer of southern Australia. Those plants that grew fastest in spring and summer were most vulnerable to sudden dehydration. Because this was an unexpected outcome, the experimental design of the field experiment only provided a semi-qualitative picture of the phenomenon, but the pattern was repeated over two contrasting seasons (Fig. 1).

There are many possible reasons for rapid dehydration. Soil water could have been depleted by profligate water use earlier in the season, a common phenomenon in crop plants (Passioura 2002). Soil water supply alone was considered unlikely to explain the effect in snow gum as the effect was clearly repeated in wetter and drier seasons. The alternatives are that root systems could have been too shallow or poorly developed, conductance of sap too slow or subject to embolism or stomata too unresponsive to water deficits to avoid catastrophic dehydration with the rapid onset of hot, dry conditions in summer. In European beech, Heath & Kerstiens (1997) suggested that total water use increased under elevated [CO2], pre-disposing trees to a greater risk of drought. In field-grown snow gums, xylem cross-sectional areas and the relative flows that they would be predicted to sustain increased rapidly around the time of sudden tree death in midsummer. As was the case for N-fertilized poplar (Harvey & van den Driessche 1999), the radial areas of xylem vessels were often larger in snow gums grown at high N (e.g. year 1; Table 1). However, considering the scale of this change, it is difficult to explain the particular vulnerability of trees at high N to dehydration relative to trees grown generally with elevated [CO2]. Nonetheless, the stimulatory effects of high N and elevated [CO2] on vessel dimensions would have theoretically raised the vulnerability of trees to both exhaustion of soil water reserves and embolism during periods of maximum transpiration in midsummer. Alder, Sperry & Pockman (1996) found that in populations of Acer grandidentatum, root xylem was more vulnerable to embolism than shoot xylem. Based on this observation, vessels at the root–shoot junction of snow gum with mean diameters approaching 60 µm would be at high risk of embolism. To establish that these wider vessels were functional, flow characteristics were further investigated by measuring the conductance of entire root systems.

There was no significant effect of N supply on the hydraulic conductance of root systems. That is, the slopes of hydraulic curves such as those plotted in Figs 2 and 4 obtained under different N treatments were similar in spite of differences in mean vessel area across N treatments, indicating either that xylem cross-sectional areas did not vary sufficiently to affect conductance or that resistances elsewhere in the soil–root pathway were dominant (Bunce & Ziska 1998; Brodribb & Hill 2000). Because axial resistance to sap flow is a small component of the total root resistance, it is likely that the radial vessel dimensions reported do not always scale closely with total hydraulic flow, as we showed for stems of snow gum in the field (Fig. 3). In coniferous species from xeric and mesic sites with contrasting fertility, effects of soil fertility on plant hydraulics seemed to be the indirect result of allometric changes (e.g. in root-to-leaf area ratios) rather than direct effects on the conducting system (Addington et al. 2006). Moreover, in Eucalyptus grandis, N fertilization did not affect shoot hydraulics (Clearwater & Meinzer 2001), whereas N caused small but significant increases in specific conductivity of stem segments of savannah trees (Bucci et al. 2006). By contrast, when comparing conductance of snow gum from the two [CO2] treatments, there was a highly significant increase in below-ground conductance which correlated with the significantly larger vessels at the root–shoot junction of CO2-enriched trees. On these grounds, we claim that roots of these trees could transport more water even though they would be more vulnerable to dehydration via embolism without tight stomatal control. Reduced stomatal apertures of snow gum at elevated [CO2] in the OTCs could mitigate this risk by improvements to water-use efficiency (Roden, Egerton & Ball 1999). Gartner, Roy & Huc (2003) reported an increase in the frequency of large-diameter vessels (without any change in the total frequency of vessels or proportion of stems given over to vessels) when Quercus ilex seedlings were grown in CO2-enriched atmospheres for 16 months. This contrasts markedly with shoots of snow gum, where the capacity of individual branches to conduct sap did not increase under CO2 enrichment, while the number of branches that comprise the shoot was responsive to [CO2] (Table 2). Branching induced by elevated [CO2] is a common observation in many species (references cited in Saxe, Ellsworth & Heath 1998) and offers a natural mechanism of regulating development and water use. We speculate that the dominance of the primary axis typical of woody root systems drives an evolutionary requirement for plasticity of vessel dimensions at the root–shoot junction.

The remaining question as to why high N predisposed trees to higher risk of dehydration, in the absence of any clear N effects on hydraulics, was pursued in a glasshouse experiment. The highly controlled and moderate temperatures were not intended to mimic field conditions, rather to establish the specific effects of N and [CO2] on plant development, allow for easier assessment of hydraulic conductance and enable sampling of entire root systems.

Low N supply suppressed shoot : root ratio (dry mass basis), particularly as plants became established under ambient [CO2] conditions, reflecting a more extensive root system as an adaptation to N depleted soils (Harley & Scott Russell 1979). The stimulatory effect of elevated [CO2] on carbon allocation to root biomass was marked in snow gum as in other woody species. Previous reports suggest that CO2 enrichment enhances fine root density more than overall root mass (Norby et al. 1999), but this was not measured in our experiment. However, the strongest effect of high soil N was to reduce root penetration into the soil profile. In ambient and elevated [CO2] conditions, respectively, the main taproot was 15–40% shorter when high N was applied. This effect, when combined with the significant increase in shoot biomass and the smaller but positive effect of high N on total leaf area, led to a disruption of the shoot : root equilibrium in trees grown in high N soils (shoot : root biomass and shoot : root length ratios in Table 2). We believe that this allometric shift away from effective root systems contributed to dehydration of field-grown trees at high N in late summer. In the glasshouse, there were no hot dry conditions or tree mortality. A similar imbalance between roots and shoots was identified when Cerrado trees from central Brazil were grown with additional N fertilizer (Bucci et al. 2006). In addition, in P. taeda L. trees from unfertilized plots, more C was generally allocated below-ground than in fertilized plots (Palmroth et al. 2006). The general relationship between C allocation above- and below-ground in other woody plants treated with elevated CO2 and fertilizer was repeated in snow gum.

The possibility that internal water transport capacity through roots further exacerbated water deficits in the canopy was investigated by calculating hydraulic conductivities. Destructive estimates of root depth after measurement of hydraulic flow allowed conductivity to be estimated by calculating suction gradients (kPa mm−1) along the entire taproot. These experiments showed that conductivities were not significantly affected by N status of the plants, as was observed in field-grown trees. This is consistent with the fact that vessel dimensions were identical between N treatments in the glasshouse.

Sapwood-specific hydraulic conductivity was also unaffected by elevated [CO2] in glasshouse-grown plants, again fitting the observation that mean vessel cross-sectional area did not increase in controlled conditions. Both rapid shoot growth and development of large late secondary xylem vessels that typified field-grown trees in summer appear to have been repressed in the constant, mild and more humid conditions of the glasshouse. Therefore, we believe that large sap conductance in roots of field-grown trees in late summer was a specific developmental response to summer conditions and could have made trees more vulnerable to embolism during days exceeding 35 °C. Prior & Eamus (2000) have also raised the possibility that death of whole shoots of eucalypts in tropical Australia might result from root or lignotuber embolism.

Constant SSC and LSC under ambient and elevated [CO2] conditions in the constant environments were seen in other species. For example, Atkinson & Taylor (1996) also found that specific conductivities were generally unaffected by elevated [CO2] in cherry and oak stems grown in glasshouses, as did Tognetti et al. (1999) in trees surrounding the natural CO2 vents in Italy. By contrast, XSCs doubled in snow gum grown at elevated [CO2]. While 67% more sapwood area in snow gum provided sufficient transport capacity to supply water to the 47% larger canopies of trees under elevated [CO2], we suggest that when root penetration was compromised, particularly at high N, more efficient delivery of sap via root xylem could only be achieved through greater XSCs. The basis of any such compensatory response is not known. Possibly, a small cohort of wide conducting vessels enhanced flow (see Tyree & Ewers 1991) without significantly affecting the mean cross-sectional area of the 300–700 vessels found in each stem. Data from droughted vines suggest that the radial dimensions of vessels influence conductivity strongly (Lovisolo & Schubert 1998), generally accounting for 20–100% of overall conductivity among woody species (Tyree & Ewers 1991). Alternatively, other resistances in the sap flow pathway could be lower in plants grown at elevated [CO2] (e.g. root radial resistance, end-wall resistance). For example, there is evidence that root radial resistance (water transport to the xylem) can change independently of axial resistance (Nobel & Alm 1993), whereas xylem end-wall resistances probably scale closely with the wall resistance that relates to vessel diameter (Sperry, Hacke & Wheeler 2005). The contrast between the minimal hydraulic changes in the aerial branches of snow gum and strong effects at the root–shoot junction of the same plants supports our view that hydraulic control is dynamic in the roots of snow gum and, together with root morphology, might be critical for survival.


Evidence presented here shows that elevated [CO2] had a very strong influence on hydraulic characteristics of trees grown under both field and glasshouse conditions. In the field, elevated [CO2] caused large secondary xylem vessels to form during growth spurts throughout summer, thus enhancing SSC. By contrast, in the glasshouse, xylem vessels were fairly constant in diameter and yet XSC increased under elevated [CO2] to supply large leaf canopies. This concurs with literature that suggests hydraulics and growth are linked through the increased requirement for water supply to larger leaf canopies (Wullschleger et al. 2002). Whether enhanced growth explains all the changes we observed in axial hydraulics, or they are the consequence of genes expressed specifically as a result of accelerated stem development (Saxe et al. 1998), is not determined in this paper.

In contrast to elevated [CO2], N appears to have accelerated shoot growth while disturbing the dynamic equilibrium of trees via changes to the shoot : root ratio. Indirect supporting evidence for this disequilibrium comes from growth data collected in the field where ‘high N’ trees under ambient [CO2] conditions had leaf canopies comparable in size to ‘low N’ trees grown in elevated [CO2], and yet the former died more rapidly during heatwaves. It is clear that canopy size per se was not a factor in sudden death, but rather the low root–shoot ratio induced by high N status. Specifically, the shallow roots and greater shoot biomass produced by snow gum in response to high N supply predisposed the canopies to rapid dehydration in summer. The hazard imposed by N fertilization was overwhelming in the field, even when the mitigating effect of ambient [CO2] on growth and hydraulics was considered. This supports the speculation of Ewers et al. (2000) that hydraulic failure occurs when previously fertilized P. taeda trees under ambient atmospheric conditions were subjected to drought.


The authors are very grateful to Jack Egerton for establishing and maintaining the field site; Wayne Pippen, Gillian Langford, James Hooker and Mohammad Masood for assistance in harvests and hydraulic measurements; Greg Joss for developing macros for image analysis; and Judy Davis for preparation of final figures.