Drought response strategies define the relative contributions of hydraulic dysfunction and carbohydrate depletion during tree mortality


Author for correspondence:

Patrick J. Mitchell

Tel: +61 3 6237 5624

Email: patrick.mitchell@csiro.au


  • Plant survival during drought requires adequate hydration in living tissues and carbohydrate reserves for maintenance and recovery. We hypothesized that tree growth and hydraulic strategy determines the intensity and duration of the ‘physiological drought’, thereby affecting the relative contributions of loss of hydraulic function and carbohydrate depletion during mortality.
  • We compared patterns in growth rate, water relations, gas exchange and carbohydrate dynamics in three tree species subjected to prolonged drought.
  • Two Eucalyptus species (E. globulus, E. smithii) exhibited high growth rates and water-use resulting in rapid declines in water status and hydraulic conductance. In contrast, conservative growth and water relations in Pinus radiata resulted in longer periods of negative carbon balance and significant depletion of stored carbohydrates in all organs. The ongoing demand for carbohydrates from sustained respiration highlighted the role that duration of drought plays in facilitating carbohydrate consumption.
  • Two drought strategies were revealed, differentiated by plant regulation of water status: plants maximized gas exchange, but were exposed to low water potentials and rapid hydraulic dysfunction; and tight regulation of gas exchange at the cost of carbohydrate depletion. These findings provide evidence for a relationship between hydraulic regulation of water status and carbohydrate depletion during terminal drought.


Changes in temperature and rainfall patterns across many forest and woodland ecosystems are thought to underlie the increasing vulnerability of tree species to drought-related mortality (Allen et al., 2010). Global analyses of forest vulnerability and drought-related mortality events across different biomes suggest that a wide range of forest types are susceptible to periods of extreme temperature and water deficit (Allen et al., 2010; Choat et al., 2012). However, in many cases, mortality rates vary among species or functional types (Pook et al., 1966; Breshears et al., 2005; Fensham & Fairfax, 2007) suggesting strong differences in drought resistance among co-occurring species. Despite major progress on understanding on how water deficit affects plant functioning (Sperry, 2000; Breda et al., 2006; Flexas et al., 2006) the controls on species survival under extreme drought remain poorly resolved and limit our ability to adequately predict future changes in ecosystem structure and function.

Recent syntheses have built on earlier work on drought physiology to describe how mechanisms of mortality are linked to plant capacity to regulate its carbon and water balance under drought conditions of differing intensity and duration (McDowell et al., 2008; Allen et al., 2010). These frameworks suggest that, for any given drought event, both plant hydraulic strategy and meteorological conditions will define the ‘physiological drought’ experienced by the plant. Physiological drought is a function of plant regulation of water-use in response to declining soil water potential and thresholds associated with hydraulic- or carbohydrate-mediated mortality. For example, plants exposed to low-intensity but long-duration droughts may maintain water status above critical water potential thresholds but deplete stored carbohydrates to lethal limits (i.e. carbon starvation). Conversely, under high-intensity drought, incapacity to regulate plant water status above critical thresholds will promote xylem cavitation and death through dehydration (i.e. hydraulic failure). In addition to these drivers, trees often sustain tissue damage from biotic and abiotic agents such as pest outbreaks (Ayres & Lombardero, 2000) and extreme temperatures (De Boeck et al., 2010) that can amplify the impacts of drought. This conceptual framework has been the focus of intense debate within the literature (Adams et al., 2009; McDowell & Sevanto, 2010; Sala et al., 2010) but has not been extensively tested with observational and experimental data.

The processes of hydraulic failure and carbon starvation are intimately linked (McDowell et al., 2011) via the balancing act required to minimize water loss and maximize carbon uptake at the leaf surface. Adjustments in stomatal aperture act to reduce transpiration in response to declining hydraulic conductance and/or reductions in leaf turgor and help plants avoid water potentials that can induce hydraulic failure (Sperry, 2000). At the whole-plant scale, the hydraulic architecture of plants appears to be well coordinated with photosynthetic capacity (Brodribb & Feild, 2000; Santiago et al., 2004; Quentin et al., 2012)., Under favourable growing conditions, higher photosynthetic capacity tends to promote higher growth rates (Poorter et al., 1990; Lambers & Poorter, 1992). Therefore, higher rates of hydraulic conductance may promote high growth rates in trees by maintaining high stomatal conductance (gs), internal [CO2] and carbon gain (Tyree, 2003). However, during protracted periods of water deficit, high hydraulic conductance may also increase plant susceptibility to cavitation (Maherali et al., 2006). Evidence for a growth vs hydraulic safety trade-off is highlighted by relationships between wood density and growth rate (Enquist et al., 1999; Muller-Landau, 2004) and cavitation safety (Hacke et al., 2001). The capacity for rapid growth (i.e. high intrinsic relative growth rates) can reduce carbon allocation to hydraulic safety (e.g. reduced lumen fraction, thicker cell walls), potentially making these species more vulnerable to hydraulic failure during drought.

Maintenance of cell turgor, which is a hydraulically mediated process, plays an important role in regulating the carbon balance of plants. Growth is particularly sensitive to changes in cell turgor and often declines before reductions in leaf photosynthesis in response to drought (Hsiao et al., 1976; Amthor & McCree, 1990) In contrast, respiration tends to be relatively insensitive to drought and does not decline proportionally in response to declining photosynthesis (Atkin & Macherel, 2009; Ayub et al., 2011). As a result, the concentration of nonstructural carbohydrates within plant tissues depends on the balance between carbon supply (i.e. photosynthesis) and carbon demand (i.e. growth and respiration). In the short term, the concentration of nonstructural carbohydrates during drought will increase if reductions in growth precede declines in photosynthesis, a response that has been observed in many species (Tissue & Wright, 1995; Körner, 2003; Ayub et al., 2011). However, if drought is prolonged, reductions in carbon assimilation and subsequent consumption of reserve carbohydrates may occur until a threshold is reached, after which plants may die of carbon starvation (McDowell et al., 2011).

As a result of these trade-offs in hydraulic and carbon dynamics, plants operate along a continuum of responses to drought, which are characterized by varying levels of hydraulic regulation to declining water availability associated with structural and physiological traits (Mitchell et al., 2008; Bartlett et al., 2012). We propose that physiological drought will vary in species with different growth and hydraulic strategies and generate water stress conditions of differing duration and intensity that are associated with mortality processes involving hydraulic failure and carbohydrate starvation. To test this, we compared the magnitude and timing of changes in water relations, gas exchange and carbohydrate dynamics in three tree species with different water use and growth characteristics exposed to a terminal drought treatment. We hypothesized that tight hydraulic regulation of water and carbon uptake reduces the risk of rapid hydraulic failure, but results in increased duration of the physiological drought and promotes depletion of nonstructural carbohydrates.

Materials and Methods

Plant material and growth conditions

Whole-plant responses to drought were studied using three species with different patterns in growth and drought response. Eucalyptus globulus (Labill.) is a fast-growing tree with profligate water use. It is an important native and plantation forest species in south-eastern Australia, and planted extensively for wood and pulp products in temperate and Mediterranean climates around the globe. Anecdotal evidence suggests that Eucalyptus smithii (R.T. Baker), which has similar wood properties to E. globulus, is less susceptible to drought than E. globulus and as a result is being trialled as a replacement for this species in some regions. Pinus radiata (D. Don) is the most widely planted pine species in the world and an important softwood plantation species. Overall, growth and water-use in P. radiata is more conservative than in eucalypt species, based on stand-level comparisons of growth and water-use (Myers et al., 1996).

One hundred seedlings of each of the three species were obtained for this experiment. Eucalyptus globulus seedlings were obtained from the Forestry Tasmania Nursery near Perth, Tasmania. Seedlings of E. smithii were obtained from seed sown and grown at the Western Australia Plantation Resources nursery, Manjimup, Western Australia. Pinus radiata cuttings were sourced from Lanoma Estate, Westerway, Tasmania. The eucalypt seedlings and P. radiata cuttings were c. 6 months and 18 months old, respectively, at the start of the drought treatment and observations of root systems during harvest suggest that all species had explored the majority of the soil volume. Seedlings/cuttings were planted into a potting mix consisting of eight parts composted pine bark to three parts coarse river sand. The eucalypt mix contained low phosphorus premium controlled-release fertilizer (N : P : K – 17.9 : 0.8 : 7.3) and P. radiata were fertilized with a standard controlled-release fertilizer (N : P : K – 19.4 : 1.6 : 5). Three hundred 250-mm diameter black pots (8 l) were numbered and filled with the potting mix to a combined weight (pot plus mix) of 6 kg and wetted to field capacity. The mix content of each pot was compressed to 6 l volume. One eucalypt seedling or pine cutting was planted into each numbered pot.

Plants were grown in two adjacent glasshouses where growth temperatures were allowed to vary as a function of outside temperature and ranged from 3 to 38°C, with a mean of 18.1°C. Half of the seedlings from each of the three species were allocated randomly to each glasshouse. The position of seedlings within both glasshouses was changed every 2 wk. There were no significant differences between glasshouses in any of the physiological parameters among the three species (see later); thus, the data were pooled for subsequent analyses.

Watering regime

Equal numbers of plants from each species were randomly allocated to one of two treatments (control/well-watered or drought). Plants in the control treatment were maintained at field capacity throughout the experiment utilizing daily drip irrigation. All control plants were watered for 30 min daily at a rate of 4 l h−1. The drought treatment was designed to achieve a gradual dry-down of the soil by applying a watering regime that replaced 80–90% of the water lost via evapotranspiration during the previous 24 h. Every 2 wk, six randomly selected plants from each species were assigned to a separate bench and used as representative samples for calculating the watering regime. On a daily basis, these plants were weighed (± 0.1 g; Sartorius BP1600S, Goettingen, Germany) between 16:00 h and 18:00 h. We estimated the total amount of water lost via evapotranspiration (including evaporation from bare soil) as the difference between the previous day's pot mass (after watering) minus the current day's pot mass. To smooth out day-to-day variation in this water loss, the actual volume of water replaced on a daily basis was calculated as a running mean of the previous 3 d. Water supply to the plants was gradually reduced until < 10 ml of water was being added to each plant on a daily basis. The drought-treated plants received water using this method for c. 60, 80 and 120 d for E. globulus, E. smithii and P. radiata, respectively. After this point, no water was added and pots were allowed to dry until symptoms of plant death were noted. In our study, plants were considered to be dead when all leaf material had turned brown, stems were brown and respiration rates in the stem were zero (as estimated by gas exchange measurements described below). To confirm our diagnosis, a subset of plants exhibiting these symptoms was rewatered and did not recover after > 1 month of rewatering.

Gas exchange

Leaf gas exchange measurements were conducted on mature, fully expanded leaves from six plants per treatment and species using a portable gas exchange system (Licor-6400; Licor, Lincoln, NE, USA). Light-saturated photosynthesis (Asat) was measured with a standard leaf chamber equipped with blue–red light-emitting diodes at a photosynthetic photon flux of 2000 μmol m−2 s−1 and [CO2] of 390 μl l−1. Leaf-to-air vapour pressure deficit during measurements varied between 1 kPa and 1.5 kPa, and leaf temperature was maintained at c. 20°C. The Asat measurements were conducted mid-morning (09:00–12:00 h) at 2–4 wk intervals throughout the course of the experiment.

Dark respiration rates (Rdark) were measured at 20°C on detached leaves (sampled immediately before measurement) from six plants per treatment that had been placed in a dark temperature-controlled room for at least 2 h before measurement. The Rdark measurements were conducted on the same day as Asat using a large transparent conifer chamber attached to the same gas exchange system (Licor 6400-05). Stem respiration or lack of (used to assess tree death) was measured on dark-adapted stems using the same chamber and gas exchange system. Stems were clamped using the middle of the conifer chamber and the gasket stem interface was sealed using Blu Tack (Bostik International, Paris, France).

Soil-to-leaf hydraulic conductance

Soil to leaf hydraulic conductance was estimated at 2–4 wk intervals to provide an estimate of the extent to which whole-plant water transport was affected by reductions in hydraulic conductance via cavitation in xylem tissues. Total soil to leaf hydraulic conductance (KP, mmol  m−2 s−1 MPa−1) was estimated as:

display math(Eqn 1)

where E is whole-plant transpiration (mmol m−2 s−1) and bulk soil water potential, Ψsoil (MPa), and leaf water potential, Ψleaf (MPa), are estimated from predawn and midday leaf water potential, respectively (Whitehead & Jarvis, 1981).

On the evening before the day of measurements, six plants from each species in each treatment were randomly selected. Each pot was wrapped in plastic bags and sealed around the base of the stem to prevent soil evaporation during the following day. Overnight, the above-ground portion of each plant was totally covered in large plastic bags to ensure that there was no nocturnal transpiration and to bring plant water potential into equilibrium with soil water potential.

Approx. 1 h before sunrise, the plastic bags covering above-ground biomass were removed. Pre-dawn leaf water potential (Ψpd) of one leaf from each plant was measured using a Scholander-type pressure chamber (PMS instruments, Corvalis, OR, USA). Transpiration (E) was estimated by weighing each of the measurement pots at midday. After the pots were weighed, a leaf from each plant was sampled for the determination of midday Ψleaf. After c. 1 h, each of the pots were reweighed. Transpiration was calculated as the weight lost over the measurement interval (making sure to account for the weight of the leaf material removed for determination of Ψleaf); thus, E could be expressed as mmol (H2O) m−2(leaf area) s−1. The height and diameter of each plant was recorded and leaf area estimated from allometric relationships developed from destructive sampling of the plants (see later for details).

Pressure-volume analysis

Leaf pressure-volume curves were generated on leaves from six control plants per species using the bench drying technique (Turner, 1988). Briefly, leaves were sampled between 07:00 h and 09:00 h from well-watered plants and rehydrated for 3–4 h in the dark. Before commencing measurements, petioles were recut under water and blotted dry. Timing of measurements of leaf weight (± 0.1 mg) and Ψleaf was done to ensure adequate points during the initial phase of rapid water loss and 3–5 points on the linear phase (post-turgor loss) of the curve. The leaf water potential at turgor loss (ΨTLP) and relative water content at turgor loss for each sample was calculated using a pressure–volume curve-fitting routine (v 5.5; K. Tu, University of California Berkeley, CA, USA) based on the template of Schulte & Hinckley (1985).

Destructive harvesting

Six plants from each species and treatment combination were destructively harvested at approximately monthly intervals from beginning of the experiment until 2–4 wk after death. These plants were used to determine relative growth rate (RGR) and develop allometric relationships used to estimate leaf area for calculating E (= 36; r2 = 0.67–0.81) and subsampled for carbohydrate analyses. Harvested plant biomass was divided into leaves, stem and roots. A sample of five to ten leaves was used for determination of specific leaf area (SLA). Leaf area was determined using a scanner and software (WinRhizo; Reagent Instruments, Quebec City, QC, Canada). Immediately after harvesting, each biomass component was placed into an oven at 110°C for 1 h to kill any biological activity. Plant material was then subsequently dried to a constant weight at 65°C.

RGR (mg g−1 d−1) was determined using the ‘classical’ method (Hunt, 2002) whereby RGR was calculated across one harvest interval using:

display math(Eqn 2)

Where ULR is the unit leaf rate; LWF is leaf weight fraction; W is plant dry weight, La is leaf area; Lw is leaf dry weight. Relative growth rates were calculated using an excel worksheet described in detail by Hunt (2002).

Carbohydrate analyses

Subsamples of oven-dried material from the destructive harvests were ground to a fine powder in a ball mill. Soluble carbohydrates were extracted from c. 20 mg of dried plant tissue in 5 ml of 80% aqueous ethanol (v : v) in a polyethylene tube. The mixture was boiled in a water bath at 95°C for 30 min, and then centrifuged at 3000 rpm for 5 min. The supernatant was collected and the pellet re-extracted once with 5 ml of 80% aqueous ethanol (v : v) and once with 5 ml of distilled water then boiled and centrifuged as before. The supernatants were reserved and evaporated (until 1–3 ml remained) in a rotational vacuum concentrator at 40°C. Total soluble sugars were determined on the supernatants following the anthrone method (Ebell, 1969). Total starch was determined on the remaining pellets after the ethanol and water extractions and assayed enzymatically using a total starch assay kit (Megazyme International Ireland Ltd, Wicklow, Ireland). Soluble sugar (SS) and starch (St) concentrations (mg g−1) were calculated as content of the measured pool divided by dry weight of the sample. Whole-plant SS and St was calculated as the sum of the weighted concentrations (concentration multiplied by the proportion of component dry mass to total dry mass) of the different biomass components (leaf, stem and root). Total nonstructural carbohydrate (TNC) concentrations were calculated as the sum of whole plant SS and St.

Statistical analysis

One-way analyses of variance (ANOVA) were performed to test for between-species differences in predrought RGR, KP and ΨTLP (< 0.05). Differences between treatments in Ψpd and gas exchange parameters (Asat and Rdark) for each species were tested using a two-way ANOVA using day of drought and treatment as variables (< 0.05). A two-way ANOVA was used to test for treatment differences in TNC, SS and St among plant organs (leaf, stem, and root) at mortality for each species. Changes in TNC, SS and St concentrations within different treatments between the first biomass harvest (predrought conditions) and the final harvest (2–4 wk after all plants were dead) were calculated (% change in concentration at mortality vs concentration before drought minus 100) to determine accumulation vs depletion of carbohydrates in each biomass component. Between-treatment differences were then examined using a two-way ANOVA (P < 0.05) using organ and treatment as factors. Student's t-tests were performed on whole-plant TNC at mortality to test for treatment differences (< 0.05). All analyses were performed in R (v2.13.0; Foundation for Statistical Computing).


Plant mortality

The mean time-to-mortality for each species, based on the complete desiccation of above-ground biomass (and associated symptoms), ranged from c. 90 d for E. globulus to c. 215 d for P. radiata (Table 1). Time-to-mortality among individuals within the same species varied from 1–2 wk for the eucalypt spp. to 2–3 wk for P. radiata.

Table 1. Summary of differences in leaf turgor loss point (ΨTLP) and corresponding relative water content (brackets), relative growth rate (RGR) and soil-to-leaf hydraulic conductance (KP) during predrought conditions and day of drought at mortality for each species (± 1 SE)
SpeciesΨTLP (MPa)RGR (mg g−1 d−1)KP (mmol m s−1 MPa−1)Day of drought at mortality
  1. Superscript letters denote groupings of species based on pairwise comparisons (< 0.05).

Eucalyptus globulus −2.22 (0.77) ± 0.04a0.024 ± 0.006a4.43 ± 0.14a90
Eucalyptus smithii −2.00 (0.88) ± 0.05b0.028 ± 0.012a5.76 ± 0.14b130
Pinus radiata −1.66 (0.87) ± 0.09c0.011 ± 0.008b2.31 ± 0.18c215

Growth rates and plant water relations

Mean total dry biomass and leaf area before the initiation of the drought was 67 g and 0.17 m−2 for P. radiata, 34 g and 0.20 m−2 for E. globulus, and 20 g and 0.13 m−2 for E. smithii, respectively. The two eucalypt species had similar RGR during predrought conditions (0.024–0.028 mg g−1 d−1; Table 1). In contrast, P. radiata had significantly lower RGR (P < 0.05; 0.011 mg g−1 d−1; Table 1). The ΨTLP showed a contrasting pattern and was highest in P. radiata (−1.6 MPa) and lowest in E. globulus (−2.2 MPa), while E. smithii was intermediate (−2.0 MPa; < 0.05, Table 1). Relative water content at turgor loss was also lowest in E. globulus (0.77; Table 1) and higher in the other two species (0.87; Table 1).

We used the time taken for Ψpd to reach ΨTLP to gauge how quickly water deficit progressed in these three species, based on the average progression of leaf water status using fitted linear functions of day of drought vs natural logarithm (Loge) (*−1) Ψpd (r2 = 0.80–0.89). Despite E. globulus having a lower ΨTLP, the time taken for Ψpd to reach this threshold ΨTLP was the shortest (44 d) of the three species. Eucalyptus smithii had a slightly higher ΨTLP, but it took 68 d for Ψpd to reach this ΨTLP. Pinus radiata had the highest ΨTLP, but spent 93 d above the ΨTLP (Fig. 1a). The decline in Ψpd post-turgor loss was very rapid in the eucalypt species. Between-tree variability in Ψpd was low until Ψpd were between −4.0 MPa and −4.5 MPa for the eucalypt species; below these values, within- and between-plant variation became very large, which was indicative of loss of hydraulic conductance across much of the canopy. All species reduced whole-plant E (as % of max E in control plants) to c. 5% at the ΨTLP while Kp (expressed as a% of max Kp in control plants) was reduced by < c. 20%.

Figure 1.

Progression of plant water relations for control (closed symbols) and droughted (open symbols) Eucalyptus globulus, Eucalyptus smithii and Pinus radiata seedlings. (a) Day of drought and mean predawn leaf water potential (Ψpd, ± 1 SE). (b) Day of drought and soil-to-leaf hydraulic conductance (KP, ± 1 SE). The start of the drought treatment is denoted by the vertical dotted line. The KP data are fitted with a linear (eucalypt spp.) or exponential decay (P. radiata,< 0.05) for E. globulus (drought; = −0.05x + 3.11, r2 = 0.88 and control; ns), E. smithii (drought; = −0.04x + 3.27, r2 = 0.92 and control = −0.04x + 5.04, r2 = 0.51) and P. radiata (drought; = 2.16 × e0.02x, r2 = 0.88 and control; = 4.00 × e−0.01x, r2 = 0.38). The leaf water potential at turgor loss (ΨTLP) for each species is shown in the left panels (horizontal dash-dot line).

Before the onset of drought, KP was higher in the two eucalypt species (4.4–5.8 mmol s−1 m−2 MPa−1) than in P. radiata (2.3 ± 0.2 mmol m−2 s−1 MPa−1; Table 1). While KP declined in the well-watered eucalypt control plants throughout the course of the experiment (Fig. 1b), the decline in KP in response to the drought treatment was more marked and declined linearly with days of drought. This rate of decline was more rapid in E. globulus and E. smithii (−0.05 mmol m−2 s−1 MPa−1 d−1 and −0.04 mmol m−2 s−1 MPa−1 d−1, respectively) than in P. radiata; in contrast to the eucalypt species, the rate of decline in KP with the progression of the drought in the pine was best described by a negative exponential function (P < 0.01, Fig. 1b). Based on these relationships and the Ψpd data, we estimated that KP became zero at c. −3.5 MPa for the eucalypt spp. and reached very low values (< 0.05 mmol m−2 s−1 MPa−1) in P. radiata at c. −2.4 MPa. P. radiata displayed comparatively tighter hydraulic regulation, whereby KP (expressed as a percentage of control KP) declined more rapidly with changes in midday leaf water potential (Ψmd) than the eucalypt species (Fig. 2). These patterns were also reflected in the rate of water loss in response to declining water availability, as indicated by the slope of the predawn Ψleaf vs Loge E, which was largest for P. radiata (2.1 mmol m−2 s−1 MPa−1) and significantly smaller (< 0.05) for the eucalypts (1.0 mmol m−2 s−1 MPa−1 and 0.8 mmol m−2 s−1 MPa−1 for E. globulus and E. smithii, respectively; Fig. 3). These data suggest that regulation of water loss in P. radiata was more sensitive to changes in soil water availability than in the two eucalypts.

Figure 2.

Relationship between mean midday leaf water potential (Ψleaf) and soil-to-leaf hydraulic conductance (KP), expressed as a percentage of control for each sampling occasion. The leaf water potential at turgor loss (ΨTLP) for each species (Eucalyptus globulus, Eucalyptus smithii and Pinus radiata) is shown as a vertical dashed line. Data are fitted with a three-parameter sigmoidal function; a/(1 + e(−(xc)/b)) (< 0.05).

Figure 3.

Relationship between mean predawn leaf water potential (Ψpd) and whole-plant transpiration (E, natural log transformed) for the three study species. Each species data set are fitted with a linear regression (< 0.01); Eucalyptus globulus (red circles); = 1.03x + 1.32, r2 = 0.88, Eucalyptus smithii (green squares); = 0.76x + 1.06, r2 = 0.86 and Pinus radiata (blue diamonds); = 2.12x + 1.18, r2 = 0.90.

Gas exchange

Predrought, Asat was highest in E. smithii (Asat = 16.1 ± 0.5 μmol m−2 s−1), intermediate in E. globulus (10.5 ± 0.6 μmol m−2 s−1) and lowest in P. radiata (3.5 ± 0.3 μmol m−2 s−1; Fig. 4). These patterns were also reflected in Rdark for the three species. Eucalyptus smithii had the highest respiration rates (1.8 ± 0.13 μmol m−2 s−1), E. globulus had intermediate rates (0.5 ± 0.03 μmol m−2 s−1) and the lowest rates were in P. radiata (0.2 ± 0.01 μmol m−2 s−1; Fig. 4). Both Asat and Rdark of the well-watered control plants declined during the course of the experiment, reflecting natural reductions in these parameters with shorter days and onset of winter. Nonetheless, drought significantly reduced Asat relative to the control treatment in all species; E. globulus reached zero (as estimated given the accuracy and precision of the gas analyser) at c. 50 d, followed by E. smithii (c. 80 d) and then P. radiata (c. 100 d, Fig. 4). The number of days of drought until plants reached a point of zero carbon assimilation was similar to the time taken to reach ΨTLP among all three species (see previous section). In contrast, for the majority of the drought (up until c. 2 wk before mortality), Rdark was not significantly different between treatments for any of the species (< 0.05). These patterns in gas exchange resulted in increased Rdark: Asat during the drought sequence in all three species (data not shown). We used the day of drought when Asat of each species reached zero as an estimate of the date when plants experienced a negative carbon balance; that is, we used this date to assume that carbon assimilation was less than whole-plant carbon utilization (Fig. 4). We found that E. globulus died c. 40 d after Asat c. 0, followed by E. smithii (c. 50 d) and P. radiata (c. 115 d). This duration of negative carbon balance was similar to the period of time droughted plants remained at zero turgor among the three species (i.e. Ψleaf < ΨTLP).

Figure 4.

Progression of leaf carbon assimilation of control (closed symbols) and drought (open symbols) seedlings of Eucalyptus globulus, Eucalyptus smithii and Pinus radiata. For each species, changes in mean light-saturated leaf photosynthesis (Asat, ± 1 SE) is shown in the left panel and mean leaf dark respiration (Rdark, ± 1 SE) is shown in the right panel. The vertical lines (dash–dot) denote estimated points along the drought sequence that net leaf carbon assimilation became zero or negative.

Changes in nonstructural carbohydrate concentrations

The depletion or accumulation of St and SS was represented as the proportional change in carbohydrate concentration at the time of mortality relative to carbohydrate concentration before the initiation of the drought treatment. For the control (well-watered) plants, there was a consistent pattern of accumulation of St in the roots of all species over the duration of the experiment. There was also significant accumulation of St in the stem and leaves in control P. radiata plants (Fig. 5, P < 0.05) and in the leaves of E. smithii. In contrast, the drought treatment promoted significant depletion of leaf St in all species at mortality, although for the two eucalypt species this depletion was largely offset by significant accumulation of SS (Fig. 5, P < 0.05). Pinus radiata showed significant depletion in the root and stem St pools and leaf and stem SS pools relative to initial conditions (Fig. 5) and a significant depletion in whole-plant TNC concentration in the P. radiata plants at mortality (−48%; Fig. 6). Furthermore, TNC concentrations were, on average, 41% lower at mortality in all biomass pools in droughted P. radiata plants relative to control plants (Table 2). However, there was no significant depletion in St or SS in the roots or stems of either eucalypt species relative to pre-drought conditions. These temporal patterns in carbohydrate concentrations indicate that the lower root St and SS in the droughted eucalypt plants at mortality were mostly a result of accumulation of these pools in the control plants rather than depletion owing to drought (Fig. 5, Table 2). Furthermore, there were no significant changes in the whole-plant TNC concentrations at mortality in the eucalypt species (Table 2).

Table 2. Carbohydrate concentration of starch (St) and soluble sugars (SS, mg g−1 DW ± 1 SE) for different organs in each species treatment at the final harvest
 LeafStemRootWhole plantWhole plant
  1. Whole plant St, SS and total non-structural carbohydrates (TNC) represent a weighted average for plant organ (based on the contribution of each organ to total biomass). Bold text denotes significant differences between control and drought treatments (< 0.05).

Eucalyptus globulus Control27.6 (6.7)68.1 (4.2)3.8 (1.0)28.5 (2.8)7.9 (1.5)38.2 (4.3)14.8 (2.0)45.6 (1.9)60.4 (2.6)
Drought4.3 (0.8)123.0 (5.8)2.1 (0.3)26.6 (3.8)2.8 (0.4)30.6 (3.5)2.9 (0.2)62.5 (4.3)65.4 (4.5)
Eucalyptus smithii Control55.3 (22.2)70.6 (5.0)2.6 (0.3)26.7 (4.5)13.5 (3.2)44.6 (6.7)20.1 (5.0)41.7 (2.9)61.9 (7.6)
Drought3.7 (0.7)103.3 (19.3)2.8 (0.6)28.3 (2.1)3.0 (0.4)27.4 (2.2)2.9 (0.3)51.0 (5.1)53.9 (4.9)
Pinus radiata Control22.4 (10.0)54.8 (3.4)11.6 (6.7)33.1 (3.3)31.0 (12.2)35.1 (2.2)19.0 (5.0)35.3 (1.7)54.3 (6.2)
Drought5.4 (1.1)33.2 (1.9)1.5 (0.9)21.0 (3.0)2.2 (0.5)47.4 (4.5)2.8 (0.4)30.8 (0.8)33.7 (0.8)
Figure 5.

Percentage change in starch and soluble sugar concentrations in root, stem and leaf components in drought (red) and control (blue) seedlings of the three study species (Eucalyptus globulus, Eucalyptus smithii and Pinus radiata). Changes in carbohydrate concentrations for each component are calculated by dividing concentration at mortality by concentration before the start of the drought treatment (−100%). Positive values denote accumulation and negative values represent depletion of carbohydrates during the experiment. Asterisks denote significant differences in carbohydrate concentrations between initial and final harvests using < 0.05.

Figure 6.

Conceptual diagram of the relationship between water status and carbohydrate depletion for the three study species (Eucalyptus globulus; red lines/circles, Eucalyptus smithii; green lines/squares and Pinus radiata; blue lines/diamonds) during drought. (a) The trajectory of predawn leaf water potential (Ψpd) in response to drought (based on fitted regression equations of the three species) and (b) percentage depletion in total nonstructural carbohydrates (TNC, weighted by components) in drought plants at mortality (relative to predrought values). Arrows extending from the three species symbols on the Ψleaf trajectories to the average estimate of mortality (vertical dashed line) represent the time over which different species remained in a state of neutral or negative carbon balance before mortality.

Relationships between water status and carbohydrates

Regulation of plant water status modifies the duration and intensity of the physiological drought and subsequent consequences for whole-plant carbon balance at mortality (Fig. 6). Based on both water relations and patterns in carbon assimilation and utilization under drought, the following relationships were observed. Eucalyptus globulus experienced a short but intense drought, as measured by the time taken for Ψpd to fall below the ΨTLP. In contrast, P. radiata experienced a longer, less intense drought; E. smithii was intermediate between these extremes (Fig. 6). Whole-plant carbohydrate status was strongly linked to these differences in duration of drought , where both eucalypt species showed no significant depletion in TNC at the whole-plant level, whereas P. radiata showed markedly less TNC at mortality compared with the well-watered control (Fig. 6).


Species responses to drought are primarily determined by the rate and degree to which plant water status is hydraulically regulated. Subsequently, this behaviour mediates the duration and intensity of the physiological stress and the processes that underpin mortality. Eucalypt species were characterized by high growth rates and water use, which intensified physiological drought through a rapid depletion of soil water and resulted in plants reaching a water status that induced complete loss of hydraulic function. In contrast, the conservative growth and water-use strategies of P. radiata reduced water loss early in the drought and reduced assimilation rate but led to significant depletion of whole plant TNC and prolonged the time to tree death. These responses represent very different strategies along the drought response continuum and support predictions of McDowell et al. (2008), whereby the drought conditions and associated physiological adaptations can alter the relative contributions from hydraulic dysfunction and the exhaustion of available carbohydrates during mortality.

Hydraulic responses to drought: the higher they fly, the harder they fall

Lower ΨTLP in the eucalypt species enabled carbon assimilation to continue to lower leaf water potentials than droughted P. radiata (Table 1, Fig. 4). Hydraulic regulation of plant water status appeared to be well coordinated with ΨTLP, with all species showing similar reductions in Kp at the ΨTLP (Fig. 2). This pattern is consistent with observed responses of stomatal conductance and leaf hydraulic conductance during desiccation (Brodribb & Holbrook, 2003). In many cases, lower ΨTLP indicates greater tolerance to drought (Bartlett et al., 2012), presumably because species may extract soil water at lower water potentials, while avoiding stomatal closure. However, under sustained drought, keeping stomata open to lower water potentials, as was the case in the two eucalypt spp., exposes species to increased risk of hydraulic failure. Accordingly, Ψpd in the two eucalypt species declined quickly following turgor loss in response to the depletion of soil water, thereby resulting in xylem cavitation and relatively rapid mortality (90–130 d). The Ψpd value at which KP became zero was estimated to be c. 3.5 MPa and accords well with field studies of E. globulus that found Kp reached zero at c. 3.7 MPa, with death occurring soon after, thereby demonstrating strong evidence for hydraulic failure at these water potentials (D. A. White et al., unpublished data). This was also supported by the dramatic increase in within-canopy and between-plant variation in leaf water potential at similar water potentials (Fig. 1) suggesting a breakdown in hydraulic connections within the canopy. In contrast, Pradiata exhibited rapid declines in KP at higher ΨTLP than the eucalypts (Fig. 3), thereby reducing gas exchange and the rate of decline in Ψpd. This more conservative drought response strategy in P. radiata resulted in more gradual reductions in Kp, which approached very low values (2% of control) at higher values of Ψpd (−2.4 MPa) after c. 130 d, yet remained alive for a further c. 90 d despite the absence of any watering. A greater sensitivity of stomata to reductions Kp in this species may act to hydraulically isolate the vascular system from the drying soil and slow the progression towards hydraulic failure and dehydration.

Tighter regulation of water-use increased carbohydrate depletion

In this experiment, strategies to regulate water-use affected both the supply of carbon through stomatal closure and reduced carbon assimilation, while simultaneously lowering the demand for carbon through reduced growth rates. The temporal sequence of leaf gas exchange, where Asat is a proxy for rate of carbon uptake and Rdark is a proxy for carbon usage, were used to characterize the response of leaf gas exchange to drought and relate these temporal dynamics to whole-plant carbon balance (Atkin et al., 2007). The point at which drought reduced leaf carbon assimilation to zero varied among species, although we acknowledge that negative carbon balance may have been reached before this point. However, Rdark was relatively insensitive to the drought in all three species, except for a period of 1–2 wk before mortality, during which respiration declined significantly. These responses are consistent with previous studies of respiratory responses to drought that show that mitochondrial respiration is often maintained under drought (Atkin & Macherel, 2009). The majority of these previous studies, however, were short-duration events of between 5 and 30 d (Atkin & Macherel, 2009). Under our extended drought (90–215 d), Rdark was generally unchanged even when leaf carbon balance was negative for substantial periods. Consequently, this ongoing respiratory demand for carbohydrates in leaves was reflected in significant depletion of leaf St among all species. These patterns in St depletion under drought supports findings that identify starch pools as important integrators of carbon supply and demand (Sulpice et al., 2009). The eucalypt species had a concomitant accumulation in leaf SS that could be attributed to the initial lag between reductions in growth and carbon assimilation (Vandoorne et al., 2012) and the maintenance of this pool for ongoing metabolic function.

The two eucalypt species and P. radiata exhibited different patterns of carbohydrate utilization during extended drought. Under well-watered conditions, the two fast-growing eucalypt species accumulated starch in roots, but not in leaves or stems. Starch accumulation did not occur under drought but whole-plant starch reserves in the eucalypts were not apparently depleted at mortality. In contrast, significant starch accumulation was observed under well-watered conditions in P. radiata but significant starch depletion was observed under drought in all plant organs. The level of starch depletion in P. radiata was very large, ranging from 54% to 60% among different plant organs and 48% for the whole plant; at mortality, starch was c. 85% lower in drought plants compared with well-watered control plants. While it is very difficult to determine whether starch depletion was sufficient to generate mortality, studies of resprouting plant species, where background concentrations of carbohydrates are generally high (Pate et al., 1990), showed lethal levels of starch depletion between 81% (Cruz et al., 2003) and 96% (Canadell & López-Soria, 1998) compared with control plants. However, starch depletion sufficient to induce mortality may be less for non-resprouting species such as Pinus, which typically do not accumulate large carbohydrate reserves. Indeed, Galiano et al. (2011) reported extremely low levels of non-structural carbohydrate concentrations in stem sapwood of dead Scots pine trees (< 20% of living trees) after an extreme drought event. Our study shows for the first time that prolonged periods of negative carbon balance during drought causes significant depletion in whole-plant TNC and provides evidence for carbohydrate starvation as a contributing process for mortality, as postulated by McDowell et al. (2008). We have shown that for significant exhaustion of carbohydrates to occur, plants must remain below or near the point of negative carbon balance while avoiding irreversible cavitation. These conditions are more likely during low-intensity/long duration water deficit and/or in species that exhibit more conservative hydraulic regulation of water status.

Future research is needed to elucidate whether duration of drought alone determines gross depletion of carbohydrates at mortality, or if carbohydrate utilization and mobilization are impeded at a threshold water status (Sala et al., 2010), suggesting an interaction between drought intensity and duration. A broader ecological question is the extent to which growth rate and related traits define the size of carbon reserves and patterns in utilization and mobilization, and therefore the vulnerability of different species. If inter-specific differences in species source-to-sink strength ratios are related to RGR, this may provide a basis for predicting how plant carbon balance is affected under drought or other stressors. Our study highlights the importance of understanding how whole-plant regulation of water status determines the exposure period and degree to which a favourable carbon balance may be compromised, which may increase plant vulnerability to additional biotic and abiotic factors.

Interplay between mortality processes and drought response strategy

Our results demonstrate that the positioning of a species along the drought response continuum are characterized to a large degree by growth and maximum photosynthetic rates, hydraulic conductance and turgor loss point. These traits influence the level of hydraulic regulation of water status (intensity) and time to mortality (duration) – the two key parameters likely to underlie the degree of hydraulic dysfunction and carbohydrate depletion during severe drought stress (Fig. 6). The ‘wait it out’ strategy of P. radiata resulted in a longer duration to mortality, but also significant depletion of carbohydrate reserves. This strategy may lead to death during drought or significantly reduce the capacity to recover from sublethal drought. Under long-duration physiological drought, trees become increasingly vulnerable to losing carbohydrate sources (i.e. leaves and branches) via leaf shedding (Galiano et al., 2011) or biotic agents such as insect defoliation (Carnicer et al., 2011). In contrast, in the eucalypt species the short duration to mortality resulted in depletion of the carbohydrate reserves in the leaves only; no depletion in storage tissues such as roots and stems was observed. However, where drought conditions induce rapid and significant losses of hydraulic conductance, but not tree death (Kursar et al., 2009), recovery processes involving growth of new leaves and water-conducting tissues (Brodribb et al., 2010) or potentially the refilling of embolized conduits (Bucci et al., 2003) may require significant utilization of stored carbohydrates. Where successive drought cycles of this kind persist, tree survival in the long term may ultimately depend on the availability of stored carbohydrate.

Differences in physiological mechanisms and the associated ecological strategies among species have broad implications for ecosystem change and may be viewed in terms of their resistance and resilience to drought and related disturbances (Westman, 1978; MacGillivray et al., 1995). In ecological terms, the ‘wait it out’ strategy of P. radiata is characteristic of resistance strategies, where survival is reliant on strong regulation of water status associated with conservative growth patterns and structural and physiological traits. However, this strategy offers little capacity for recovery once critical carbohydrate thresholds have been exceeded. Alternatively, the eucalypt species typifies a resilience-type strategy, where rapid loss of hydraulic function means that they succumb to drought rapidly. However, under non-lethal conditions (e.g. low intensity and/or short duration events), plants have the potential to recover by building new hydraulic pathways and leaves through utilization of remaining carbohydrate reserves. The differential impacts of drought on species and ecosystem dynamics may be better understood by elucidating the role of ecological strategy in underpinning plant responses to droughts of differing intensity, duration and frequency.


We acknowledge the assistance of Dale Worledge, David Page and Dominik Laschinger with biomass harvesting and Renee Smith and Kaushal Tewari with carbohydrate analyses. Funding was provided by the Australian Government's Department of Agriculture, Forestry and Fisheries. Michael Battaglia, Jenny Carter and Victor Resco de Dios provided useful feedback and comments on earlier versions of this manuscript. We also thank the reviewers for their helpful comments and feedback.