• Open Access

Convergence of tree water use and hydraulic architecture in water-limited regions: a review and synthesis


Correspondence to: Melanie Zeppel, Department of Biological Sciences, Macquarie University, Sydney, NSW 2109, Australia.

E-mail: Melanie.Zeppel@mq.edu.au


Global vegetation models are used to estimate water and carbon fluxes in current and future climates. To accurately estimate these fluxes, it is crucial to incorporate tree processes, such as transpiration. Some models accurately predict fluxes in well-watered conditions; however, our ability to predict responses of trees when water availability is limited remains restricted. Including mechanistic responses of trees during drought in models will improve estimates of water and carbon fluxes.

Estimating water fluxes over large spatial scales may be calculated by combining (1) remotely sensed estimates of evapotranspiration with (2) knowledge of whether tree water use for a particular forest type or plant functional type follows universal scaling rules. There is little knowledge of whether universal scaling rules apply to water-limited ecosystems.

This review examines ‘convergence’ in relationships among tree water use, leaf area, and tree size, using Australian broad-leaved evergreen vegetation as a case study. Broad-leaved evergreen is a plant functional type commonly used in global vegetation models. If convergence is observed among leaf area and water use relationships for different species within this plant functional type, this would provide a powerful tool for scaling ecohydrological processes. This work tests the hypothesis that tree water use (Q) converges along a common relationship with leaf area for a continent-wide range of species (n = 21), spanning a 100-fold difference in size across broad-leaved evergreens, including Acacia, Corymbia, and Eucalyptus as well as Callitris.

Remarkably, the slope of the relationship between tree water use and leaf area was similar for the broad-leaved evergreen genus, Eucalyptus, despite different slopes in relationships among diameter at breast height, leaf area, sapwood area, and Q. Realistic modelling of water and carbon fluxes requires an understanding of physiological mechanisms influencing Q, for each plant functional type, and for these mechanisms to be incorporated into vegetation models. Copyright © 2013 John Wiley & Sons, Ltd.


Global vegetation and ecophysiological models are being used to estimate water, energy, and carbon fluxes in current and future climates (Bonan, 2008; Medlyn et al., 2011a), and it is important to include mechanistic responses of trees across the range of conditions that trees experience. Some models can accurately predict tree processes in well-watered conditions; however, our ability to predict the mechanistic responses of trees during low water availability remains limited (McDowell et al., 2011). Knowing the mechanistic responses of trees to low water availability and drought is essential to accurately incorporate these mechanisms into process-based ecophysiological models and global vegetation models (Adams et al., 2011). Doing so will allow improved estimates of water and carbon fluxes in future climates (Bonan, 2008; McDowell et al., 2011).

One such process is evapotranspiration. A simplified water budget may be described where evapotranspiration is the sum of soil evaporation, tree transpiration, stem flow, and canopy interception, plus run-off and deep drainage of water (Figure 1). Tree water use is often the largest component of the water budget; meaning, to quantify the water budget of a site, we need accurate estimates of tree water use (Eamus et al., 2006). Understanding and accurately quantifying tree water use as part of the water budget is important for weather forecasting, understanding climate change, hydrology, ecosystem function, and forestry production (Pieruschka et al., 2010; Asbjornsen et al., 2011). This is particularly true under changing climates because climatic variables that have a strong influence on tree water use, such as atmospheric CO2 concentrations ([CO2]) and precipitation, are often changing beyond the limits that have been previously studied.

Figure 1.

Schematic diagram representing the water budget of a water-limited forest within a (a) dry year and (b) wet year. Evapotranspiration is the sum of understorey evaporation (including soil evaporation and transpiration of understorey vegetation such as grasses and shrubs), tree transpiration, stem flow and canopy interception, plus run-off and deep drainage of water data taken from a remnant forest in New South Wales, Australia, using methods described in Zeppel et al. (2008a). Canopy height was ~14 m and tree roots were measured at depths >20 m.

An understanding of tree water use, which returns water back to the atmosphere, is vital to understand water and energy cycles. In addition, knowledge of hydraulic architecture and ecohydrology is important for quantifying the water budget of a site (Wullschleger et al., 1998; Grigg et al., 2008; Bleby et al., 2009; Yunusa et al., 2010b; Yunusa et al., 2011). Quantifying water use as part of the water budget is important for managing catchments that supply water to millions of people, such as the Murray–Darling Basin, and Catskill and Delaware catchments for New York City (Eamus et al., 2005; Chiew et al., 2009). This review examines plant–water relations across a variety of ecosystems, particularly in water-limited and nutrient-limited regions, using Australian trees as a case study.

Broad-leaved evergreen sclerophyllous trees such as those in Australia have evolved in water-limited and nutrient-limited conditions, similar to other sclerophyllous trees found in other regions such as South Africa (Palmer et al., 2008), the Mediterranean Basin (Martinez-Vilalta et al., 2003) Portugal (Paço et al., 2009), Spain, Chile, and California (Jacobsen et al., 2009). Sclerophyllous broad-leaved evergreen vegetation is hard-leaved, with adaptations to water and nutrient limitation such as deep roots, thick dense leaves and sunken stomata (Eamus et al., 2006) If there are a limited number of physiological solutions to balancing the trade-off between gaining carbon and losing water (Meinzer, 2003), then tree physiological processes from water-limited and nutrient-limited regions – such as those found in Australia – may inform mechanistic processes from other water-limited regions of the world. To estimate water fluxes using Dynamic Global Vegetation Models for water-limited and nutrient-limited regions, it is useful to examine whether relationships between tree water use and leaf area in different plant functional types, such as broad-leaved evergreen trees from these regions follow similar patterns of functional convergence, similar to that reported for a tropical rainforest (Meinzer, 2003).

To model water, carbon, and energy fluxes across large (regional or continental) spatial scales, quantifying tree water use of a particular plant functional type is required. Identifying convergent relationships in plant traits can quantify the water use of a particular plant functional type, providing robust tools for scaling ecophysiological processes and insight into the trade-offs between carbon accumulation and water loss – which is of particular interest for modellers (Meinzer, 2003; Enquist et al., 2007; O'Grady et al., 2009). Functional convergence between water use and leaf area (LA) would enable significant ecohydrological insights and allow modellers to scale up water fluxes. To achieve this, ‘functional convergence’ in relationships among tree water use sapwood area (SA), SA, LA, and tree size across species is examined. Functional convergence is defined here as the process whereby cooccurring species converge along a similar relationship between two different physiological properties, such as tree water use and diameter at breast height (DBH), and have the same rate of water use for trees of a given size (Meinzer, 2003).

Across a range of ecosystems, co-occurring species had similar rates of tree water use per unit leaf area (Hatton et al., 1998). However, Hatton et al. (1998) state that this pattern was based on a small number of species and sites (three or four). To test whether functional convergence in tree water use exists, I tested the hypothesis that the slope of the relationship between tree water use and leaf area is the same regardless of conditions across sites. A wide range of species across multiple genera, spanning a 100-fold range in tree size in various ecosystems, were tested to determine if we may scale-up water fluxes across large spatial scales. Understanding the transpiration of forests across spatial scales is becoming increasingly important as water resources vary with increasing intensity of droughts and floods.

This synthesis of tree water use in water-limited regions builds on a number of previous reviews (Calder, 1992; Wullschleger et al., 1998; Doley, 2004; Whitehead and Beadle, 2004; Asbjornsen et al., 2011). Recent research from studies conducted across a range of species and ecosystems is synthesized. Finally, to examine whether we may predict water fluxes over large spatial scales, I test the hypothesis that relationships among tree water use and leaf area are the same across eucalypt species. The aims of this paper were to test whether functional convergence occurs in the relationships between tree size, leaf area, sapwood area, and water use in sclerophyllous vegetation across sites.


The presence of functional convergence in plant species can provide a useful tool for scaling physiological processes to a landscape, regional, or continental scale (Hatton et al., 1998; Enquist, 2002; Meinzer, 2003; O'Grady et al., 2009). This can be done in conjunction with remotely sensed estimates of evapotranspiration, by allowing ecosystem-scale estimates of carbon and water fluxes. Identifying functional traits and convergence of plant forms allows an improved understanding of community adaptation to gradients in water availability and evaporative demand. This in turn enables improvements to emerging models of water fluxes across continental scales. For example, if the relationship between leaf area index (LAI) and water use may be accurately predicted for each group of plant functional types, with one such type being ‘mesic woodlands’, and if we are able to estimate LAI from remotely sensed data (Myneni et al., 2002; Hill et al., 2006; Palmer et al., 2008), then we may be able to predict water fluxes across large regions containing this plant functional type. This assumes that the water use within that functional type has ‘converged’ across species and that universal scaling rules apply (Meinzer et al., 2001). Thus, determining scaling relationships for such types will allow improvements in predicting large-scale water and carbon fluxes. Functional convergence among broad-leaved evergreen trees is now discussed, using Australian vegetation as a case study.


Convergence may be defined, following Meinzer (2003), as the phenomenon where the relationship between two plant traits (e.g., Figure 2, tree water use and leaf area) for two or more species converge to form a single equation (see Table 1 and supplementary table for relationships among tree size, LA, and sapwood area). Convergence has been observed across relationships influencing water relations, e.g. wood density and LA to SA relationships, stand water balance, and LAI (O'Grady et al., 2011). However, the presence of convergence may depend on the water availability of the site (Pickup et al., 2005; Yunusa et al., 2010a), reflecting different abilities to maximize available resources and achieve ‘ecohydrological optimality’, or use most of the available water within a site (Eagleson and Tellers, 1982). Importantly, there has been no robust test, following Hatton et al. (1998) who used three or four points per relationship, of whether the relationship between whole tree water use and LA converges across species. If this relationship is the same for all species, then there are significant ecological, physiological, and ecohydrological implications. For the first time, the presence of convergence along a common relationship between whole tree water use and LA is tested for tree species across a wide range of sites and species within Australia (Table 1).

Figure 2.

The relationship between leaf area (m2) and diameter at breast height (DBH, mm). Data presented show species with data for > 3 trees, for (a) all genera and (b) eucalypts only. The inset in (b) shows a close-up of trees with DBH to 250 mm. Lines represent linear regressions for species with sufficient data. Data obtained from references in Table 1.

Table 1. Tree water use and tree characteristics including the method used to determine tree water use, height (m) diameter at breast height (cm), leaf area index (LAI, m2 m−2) or leaf area (m2) and tree (L day−1), or stand water use (mm day−1) for remnant forests and plantations in both water-limited and water abundant sites across Australia.
SpeciesMethodHeightDiameterLAI or leaf areaSapwood areaTree water use or stand water useReference
  1. Method used to obtain tree water use estimates is indicated by compensation heat pulse method (CHPM), heat ratio method, (HRM) or ventilated chamber (VC). This table represents a selection of studies, and does not describe all studies. Where stand water use is given (mm day−1) in some instances, the range of tree diameters (mm) used to estimate stand water use are given.
Acacia aneuraCHPMn/a120 mm5·7 m257 cm214 kg day−1(O'Grady et al., 2009)
0·49 mm day−1
Angophora bakeriCHPM14 m270 mm1·3–1·9 m2 m−232–127 cm20·4–1·9 mm day−1(Zeppel et al., 2008b)
Callitris glaucophyllaCHPM14 m350 mm1·0–1·2 m2 m−2225 cm2150 L day−1(Zeppel et al., 2006)
Eucalyptus albensCHPM<23 m25 mmn/an/a21 L day−1(Eberbach and Burrows, 2006)
Eucalyptus macrorhyncha17 m24 mm20
Eucalyptus rossii – ridgetop25 mm10
Eucalyptus sideroxylon26 mm27
Eucalyptus blakelyiCHPM6 m60 mmn/an/a800 ml h – pesticide treatment(Cunningham et al., 2009)
10 m220 mm646 ml h control
Eucalyptus camaldulensisCHPMn/an/an/an/a0·30 ± 0·2 mm day−1(Doody et al., 2009)
Eucalyptus largiflorens0·25 ± 0·1 mm day−1
Acacia stenophylla0·27 ± 0·1 mm day−1
Eucalyptus camaldulensis: clayCHPMn/a0·51 ± 0·21·1 to 1·2 m2 m−2230·7 to 3·01 mm day−1(Yunusa et al., 2010a, 2010b)
Eucalyptus macrocarpa: clay0·44 ± 0·1 140·95 to 4·19 mm day−1
Eucalyptus camaldulensis: sand0·34 ± 0·20·9 to 1·0 m2 m−2160·46 to 0·69 mm day−1
Eucalyptus macrocarpa: sand0·32 ± 0·1 110·07 to 0·26 mm day−1
Eucalyptus capillosaHRM13–18 m<150 mm0·66 m2 m−230–35 cm2Max 0·9 mm day−1(Mitchell et al., 2009)
Eucalyptus cladocalyxCHPMn/a120 mm60 m275 cm255 L day−1(Benyon et al., 2001)
Eucalyptus occidentalis160 mm35 m2100 cm260 L day−1
Eucalyptus spathulata230 mm155 m2330 cm2275 L day−1
Eucalyptus leucoxylon110 mm25 m260 cm235 L day−1
Eucalyptus crebraCHPM14 m650 mm1·0–1·2 m2 m−2250 cm2120 L day−1(Zeppel et al., 2006; Zeppel and Eamus, 2008; Zeppel et al., 2008c)
Eucalyptus globulusVCn/an/a17n/a37(Greenwood and Beresford, 1979)
Eucalyptus globulusCHPMn/a  n/a (Sudmeyer and Simons, 2008)
Thinning treatment  
1251·78 m20 mm month−1
4000·78 m40 mm month
8000·49 m60 mm month
Eucalyptus globulusCPHM7 m7·3 cm3·0n/a0·09 kg s−1 m−2(O'Grady et al., 2008)
Eucalyptus globulusCHPM18·74·8–14·71·0–3·7n/a0·4 mm day−1–1·9 mm day−1(Forrester et al., 2010)
Eucalyptus globulus* groundwater depth <3 mCHPM12–17 m0·14–0·18 m3·9–7·1n/a4 mm day−1(Benyon et al., 2006)
Eucalyptus globulus* groundwater depth >3 mCHPM140·16 m3·5 m2 m−2n/a2·1 mm day−1(Benyon et al., 2006)
Eucalyptus grandisCHPM21·423·71·2–1·8 m2 m−2n/a<1·5 mm day−1(Vertessy et al., 2000; Feikema et al., 2010)
Eucalyptus grandis13·018·7
Eucalyptus camaldulensis18·627·8
Eucalyptus kochii – uncut, aquiferCHPMn/an/a18 m2n/a20–25 L day−1(Wildy et al., 2004)
Eucalyptus kochii – uncut no aquifer8–10 m210–17 L day−1
Eucalyptus kochii – coppiced4–7 m29–15 L day−1
Eucalyptus marginata – old growthHRM23–32 m> 0·6 m1·5 ± 0·15175–310 cm20·5–0·9 mm day−1(Macfarlane et al., 2010)
Eucalyptus marginata – regrowth 13–17 m0·1–0·3 m2·1 ± 0·2620–105 cm20·9–1·8 mm day−1
Eucalyptus miniataCHPM15 mn/a0·5 to 1·0 m2 m−2 Data not shown(Kelley et al., 2007)
Eucalyptus tetradonta Stand water use shown only.
Melaleuca viridiflora0·5 to 1·8 m2 m−2
Eucalyptus parramattensisHRM14 m11–41 cm1·3–1·9 m2 m−250–331 cm20·4–1·9 mm day−1(Zeppel et al., 2008b)
Eucalyptus regnansCHPM44–65 m395–89345–404 m2209–843 cm2 (Vertessy et al., 1997)
Eucalyptus rossii – ridgetopCHPM17 m0·25 mn/an/a10 L day−1(Eberbach and Burrows, 2006)
Eucalyptus sideroxylon – ridgetop0·26 m27 L day−1
Eucalyptus sieberi      (Roberts et al., 2001)
14-year-old plotCHPMn/a0·13–18 m3·6 m2 m−234–73 cm22·2 mm day−1
45-year-old0·15–57 m4·0 m2 m−219–165 cm21·4 mm day−1
160-year-old0·31–1·12 m3·4 m2 m−245–377 cm20·8 mm day−1
Eucalyptus camaldulensisCHPMn/a180 mm47 m2135 cm287 kg day−1(O'Grady et al., 2009)
Corymbia opaca150 cm281
Eucalyptus victrix380 mm60 m2157 cm251
Eucalyptus capillosaHRM13–1840–400 mm0·66 m2 m−25–35 cm20·6–0·8 mm day−1(Mitchell et al., 2009)
Eucalyptus urophyllaCHPM16–1885–95 mm1·1–2·98 m2 m−25–9 cm21·40–1·53 mm day−1(Morris et al., 2004)
Stand water useCHPMn/an/aTree belt experiment.n/a595 mm year−1(White et al., 2000)
Eucalyptus leucoxylon1·9 mm day−1
Eucalyptus saligna2·9 mm day−1
Eucalyptus camaldulensis1·9 mm day−1
Eucalyptus platypus0·8–1·2 mm day−1

Statistical analyses were conducted using R version 2.11.1 (R Development Core Team 2010). Data were tested for homogeneity of variances, and then the slopes and intercepts were compared for all species using analysis of co-variance (ANCOVA). Nonlinear regressions produce higher R2 values and have been reported for relationships described in Figures 2, 3, and 4 (Vertessy et al., 1995); however, ANCOVA assumes linear relationships, so for this analysis, linear relationships were used. Species with three or less points per species were excluded from the analysis. Data for DBH, SA, LA, and tree water use (Q) were obtained from publicly available literature from the Institute for Scientific Information, Web of Science, and industry reports (Cook et al., 2008). Data for each species were taken from different sites where available, to incorporate differences in microclimate and water availability. Data for very large trees (DBH > 800 mm) such as Eucalyptus regnans (F. Muell) were specifically obtained to examine relationships across a very wide range of tree sizes.

Figure 3.

The relationship between sapwood area (m2) and diameter at breast height (DBH; mm). Data presented show species with data for >3 trees, for (a) all genera and (b) eucalypts only. The inset in (b) shows a close-up of trees with DBH to 250 mm. Lines represent linear regressions for species with sufficient data. Data obtained from references in Table 1.

Figure 4.

The relationship between leaf area (m2) and tree water use (L day−1). Data presented show species with data for >3 trees, for (a) all genera and (b) eucalypts only. The inset in (b) shows a close-up of trees with leaf area to 150 m2. Lines represent linear regressions for species with sufficient data. Data obtained from references in Table 1.

ANCOVA showed that there were different relationships among SA, LA, Q, and DBH for different genera (Acacia, Eucalyptus, Corymbia, and Angophora, P < 0·05; Figures 2(a), 3(a), 4(a)). Within eucalypts, ANCOVA indicated that species differed in the slope of the relationship between sapwood and DBH (P = 0·03; Figure 3(a)). Similarly, the slope between LA and DBH had a different trend (P = 0·09; Figure 2(a)), and the slope between tree water use and DBH differed among eucalypt species (P = 0·01).

Despite the different slopes between DBH and LA, DBH and SA, and DBH and Q for each species, it is remarkable that the slope of LA versus Q was not significantly different across eucalypt species (P > 0·05; Figure 4). It is important to note that this relationship was not significantly different both when data from Eucalyptus regnans (the largest species) were included, and also when excluded (P > 0·05). Curiously, although the slope of the Q versus LA relationships was not different, the intercept varied among species.

Hatton et al. (1998) proposed that, within a site, tree water use per unit LA is independent of species, such that co-occurring species, with the same supply of water and demand by the atmosphere, come to similar solutions to optimize the balance between water use and LA. This hypothesis has significant implications for the estimation of large-scale water fluxes. However, few studies were available for testing this hypothesis in 1998. Since then, numerous tree water use studies have been published, allowing more robust tests including a higher number of species. This hypothesis was tested across numerous sites, testing whether relationships among Q and LA were similar across a range of eucalypt species spanning a 100-fold difference in size. Although there are different relationships among DBH and LA (Figure 2), DBH and SA (Figure 3), and DBH and Q, the slope of Q and LA was not significantly different for eucalypts (Figure 4; Supplementary table). This result provides further evidence using a much larger dataset than Hatton et al. (1998) and Zeppel and Eamus (2008) who used three or four, and two species, respectively.

The LA versus Q relationships used in this analysis were measured across a range of sites, so the differences in the intercept, among species, may reflect different nutrient and water availability rather than inherent genetically based differences. Interestingly, O'Grady et al. (2011) conducted a stand scale analysis that reported that the slope of the relationship between LAI and stand water use was not different when groundwater was accounted for. Further, the intercept also differed, as in the present study. Thus, the differences in intercept may reflect differences in microclimate or groundwater availability (O'Grady et al., 2011). Therefore, the differences in intercept among species do not definitively reject Hatton's hypothesis. Despite this caveat, these results are important as they suggest that the rate of increase in water use as LA increases is constant among species, consistent with Hatton's hypothesis. This relationship suggests that the rate of increase of water use as LA increases is the same across all eucalypts measured.

Perhaps certain intrinsic properties of leaves of all eucalypt species are similar, despite different hydraulic architecture (leaf area per tree size) or water relations (diurnal curves of stomatal conductance per leaf area). It is well known that different species may have different leaf mass area, SLA, diurnal leaf water potential (ψl), and stomatal conductance diurnal curves (O'Grady et al., 2009; Yunusa et al., 2010a). Despite these documented differences, results indicate that the increase in the rate of water use as LA increases is similar across eucalypt species. This raises the question of whether some property inherent across the sclerophyllous leaves of eucalyptus trees leads to the same increase in the rate of water use as LA increases. Further, if different co-occurring eucalypts, which experience the same weather conditions, exhibit different rates of stomatal conductance (expressed per leaf area), how do they obtain similar rates of tree water use per day? This may be due to a combination of the following reasons: (1) regulation of hydraulic architecture, such as Huber value (sapwood area : leaf area, Whitehead, 1998); (2) differences in daily patterns of stomatal conductance patterns; and (3) decreases in sap velocity as tree size increases.

First, Huber values have been reported to decrease as trees increase in size (McDowell et al., 2002; Forrester et al., 2010). Therefore, as LA of a tree increases, the sapwood area may not increase at the same rate, giving less conducting area. Second, different eucalypts exhibit different diurnal patterns of stomatal conductance and sap flow. These different patterns in diurnal stomatal conductance or water use strategies (such as high morning rates and low afternoon rates) may compensate for the differing instantaneous rates of stomatal conductance. Different species also exhibit different responses of stomatal conductance to D. Further, as tree size increases, sap velocity may decrease in eucalypts along a size gradient, leading to nonlinear increases in tree water use as tree size increases (Forrester et al., 2010). Finally, further data are required to test these hypotheses across a wide range of sites including both groundwater-dependent and non-groundwater-dependent sites.

Similar to this finding, other studies have shown that the relationship between tree size and LA (Kelley et al., 2007; Zeppel and Eamus, 2008), or tree size and sapwood area differs across species (Pekin et al., 2009) (but c.w. Meinzer et al., 2001). However, previous studies conducted at one site showed that a common positive relationship may exist between tree size and water use (O'Grady et al., 1999), or stand basal area and tree water use (Kelley et al., 2007). A study of four eucalypt species in a plantation reported significant differences in transpiration (tree water use per leaf area, L m−2 day−1) (Benyon et al., 2001). However, these results were based on three trees per species, and the species with a significantly lower tree water use per leaf area was hypostomatous (stomata on one side of the leaf). In addition, the plantation had not reached canopy closure. Thus, it is possible that convergence between leaf area and tree water use is restricted to sites that have reached canopy closure.


Functional convergence exists at the xeric end of the water availability spectrum (O'Grady et al., 2009; Yunusa et al., 2010a), with divergence at the high water availability end (Pickup et al., 2005). At sites with low water availability in Western Australia, Mitchell et al. (2008)found convergence of strategies in water use for species growing on sites having low soil depth and therefore limited water availability. In contrast, there was more variation in the strategies adopted at sites with deep soils and higher water availability. Similarly, Yunusa et al. (2010a) also found convergence of strategies of tree water use on water limited soils but not on soils that experienced higher water availability. This demonstrates the importance of considering water relations across a wide range of spatial scales, from leaf to whole tree, from stand to landscape scale (Meinzer et al., 2010) as well as across temporal scales, including across water availability gradients (wet vs dry years). In addition to considering water relations across spatial and temporal scales, predicting future water fluxes requires consideration of water relations across a range of climatic gradients.


Climate change is predicted to increase temperatures and [CO2], as well as change the timing, frequency, and distribution of rainfall. This is expected to occur in conjunction with weather extremes, such as longer and hotter heat waves, and a more intense hydrological cycle, meaning large rain events and longer periods between rain events (Hughes, 2003; Cullen and Grierson, 2007; Hennessy et al., 2007; Battaglia et al., 2009; Steffen et al., 2009; CSIRO, 2011). Each of these environmental variables influences tree water use (Eamus et al., 2006; O'Grady et al., 2009; Zeppel et al., 2012). Increased temperatures (within normal physiological limits) generally lead to higher water use, and elevated [CO2] generally leads to increased leaf area and/or increased ‘water savings’, as stomatal closure drives increased water-use efficiency of trees, when all else remains equal (Eamus and Jarvis, 1989). However, a recent study showed that increased water use due to increased leaf area offset the water savings and trees reached drought-induced mortality more quickly under elevated CO2 compared with ambient CO2 (Zeppel et al., 2012). Clearly, decreases in rainfall will reduce tree water use, again, when all else remains equal. For a thorough discussion on the impact of climate change variables on the forests of Australia, see Medlyn et al. (2011b). A challenge for tree physiologists quantifying tree water use in future climates is incorporating the various interactions between physiological processes, which are influenced by water availability, [CO2], and temperatures (Calfapietra et al., 2010; Medlyn et al., 2011b).


Undoubtedly, complex ecophysiological modelling is required to incorporate the range of physiological processes that influence tree water use (Simioni et al., 2009), yet our current models are based on an incomplete understanding of these processes (Kirschbaum, 2005). Thus, to accurately model tree water use, an improved understanding of key physiological fluxes is required. Experimental studies have reported the impact of [CO2] on growth and water relations on Acacia (Atkin et al., 1999; Roden et al., 1999), Eucalyptus (Berryman et al., 1994; Silberstein et al., 1999; Atwell et al., 2007; Atwell et al., 2009; Barton et al., 2010; Ghannoum et al., 2010a; Ghannoum et al., 2010b; Zeppel et al., 2011b), Pinus (Conroy et al., 1990a; Conroy et al., 1990b) and mangrove species (Ball et al., 1997). However, these experiments often last less than 1 year, are sparse for many plant types (Hovenden and Williams, 2010), and often provide contradictory results (see Medlyn et al., 2011b for a summary of these experiments). For example, water fluxes have been reported to decrease under elevated CO2 in some eucalypts (Barton et al., 2012) but not others (Ghannoum et al., 2010b). In addition, water fluxes at night increased under elevated CO2, opposite to daytime water fluxes, when water was abundant (Zeppel et al., 2011b; Zeppel et al., 2012). Further, experimental results on growth rates under [CO2] have also varied with increases in some experiments (Atkin et al., 1999; Ghannoum et al., 2010a) and decreases in others (Atkin et al., 1999). These different conclusions may result from differences in species, experimental conditions, and study length. Whole tree and forest water use will depend on the combined effect of [CO2] on water fluxes, leaf area, and growth rates, and current evidence is equivocal. Therefore, further experiments are needed, particularly on species that cover large areas, or are commercially important. Additionally, we need to understand plant responses to strong climatic drivers of forest water use, such as drought and heat waves.

Few studies report the impact of heat waves, and studies on mechanisms leading to drought mortality often focus on one or two species within a biome (Adams et al., 2009; Zeppel et al., 2012), but see West et al. (2012). Thus, comparisons of how multiple species respond to drought-induced mortality are needed. Experiments that quantify how mature, water-limited forests will respond to changes in [CO2] are also needed. Additionally, there is strong evidence that the timing of precipitation is changing, and extreme precipitation events are occurring more frequently (Huntington, 2006; CSIRO, 2011; Smith, 2011). Therefore, it is vital to develop an understanding of how vegetation will respond to altered timing of precipitation, as rainfall becomes more extreme and seasonality of rainfall changes. Further, transpiration transfers water from the soil back to the atmosphere, with a significant impact on albedo, radiation reaching the understory, and precipitation of the local and regional climate. This means that an understanding of transpiration under climate change is important to understand feedbacks on local and regional climates (Bonan, 2008). Thus, the large gap in our understanding of climate change processes on tree physiology needs to be addressed before physiological models can accurately predict tree water use under future climates (Raison et al., 2008; Medlyn et al., 2011b).

This study has highlighted a number of research gaps. It is noteworthy that many continents contain woodlands and savannas on water-limited and nutrient-limited soils, including regions of South Africa, California, Chile, and the Mediterranean Basin. However, much of the research, which informs vegetation models, is based on trees from regions of North America and Europe, which are less water-limited than vegetation from geologically older soils in South Africa and Australia (Eamus et al., 2006). Therefore, quantifying mechanistic processes within water-limited and nutrient-limited regions remains a key research gap, which, if filled, may be used to improve emerging vegetation models (Zeppel et al., 2011a).

Clearly, further experimental research is needed to test the effects of elevated [CO2] on mature forests (Raison et al., 2008). Mechanisms leading to drought mortality and the role of linkages between hydraulic failure, carbon starvation, and infestation also remain unresolved (McDowell et al., 2008; McDowell, 2011; Zeppel et al., 2011a). Further, the seasonality of rainfall is projected to change (IPCC, 2011), yet there remains a paucity of research on how fluctuations in water availability will influence forest processes. Grassland experiments on the impact of altered precipitation timing in the Northern Hemisphere have demonstrated key changes in physiological processes (Knapp et al., 2002; Knapp et al., 2008). However, the influence of changed timing of precipitation on trees remains unresolved, and experiments that will provide information on tree growth and water relations under changing precipitation regimes are needed. Information on (1) tree responses to extreme precipitation and (2) mechanistic processes leading to drought mortality is needed to improve the representation of physiological processes in vegetation models.


The presence of common scaling relationships, or functional convergence between tree water use and leaf area across species, when combined with remote sensing products, may allow for improved estimates of water fluxes across regional or continental scales. Results showed that (1) despite different relationships among DBH, LA, SA, and Q for a variety of species, the relationship between LA and tree water use had the same slope for all eucalypts tested, and (2) a wide range of eucalypts across Australia have the same slope between LA and Q implying that for a given leaf area, all eucalypts use the same volume of water. This implies that there are common properties, trade-offs, and constraints inherent across Eucalyptus species. This convergence of hydraulic traits can provide potentially powerful tools for spatially scaling water fluxes in forested ecosystems, in conjunction with large spatial estimates of evapotranspiration.

Because we cannot assume that transpiration in water-limited regions is the same as water-abundant regions, an improved understanding of these mechanistic processes will inform emerging global vegetation models, allowing improved predictions of water, carbon, and energy fluxes in future climates.


I am grateful to Derek Eamus for encouraging and initiating this review and editing earlier versions, and to Jim Lewis, Belinda Medlyn, Martin DeKauwe, Tim McVicar, and David Tissue and two reviewers for providing useful insights and suggestions that have much improved earlier versions of this manuscript. This study is funded by Australian Research Council Discovery Project (DP0877722), Discovery Early Career Researcher Award (DE120100518), and a Macquarie University Research Fellowship.



Vapour pressure deficit (kPa)


diameter at breast height (mm)


soil-to-leaf hydraulic conductance


leaf area (m2)


leaf area index (m2 m−2)


sapwood area (m2)


leaf water potential (MPa)


specific leaf area (cm2 g−1)

stand water use

sap velocity ∗ sapwood area of all trees in the stand (mm day−1, Table 1)


tree water use, sapwood area ∗ sap velocity, (L day−1 or kg day−1). Tree water use, Q, is defined as the volume of water flowing out of the tree. Transpiration Et is Q per unit leaf area.