Plasticity in maximum stomatal conductance constrained by negative correlation between stomatal size and density: an analysis using Eucalyptus globulus



    Corresponding author
    1. Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK,
    2. Faculty of Agriculture, Food and Natural Resources, University of Sydney, Sydney, New South Wales 2006, Australia and
    Search for more papers by this author

    1. School of Biological Sciences and Biotechnology, Murdoch University, South Street, Murdoch, Western Australia 6150, Australia
    Search for more papers by this author

    1. Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK,
    Search for more papers by this author

P. J. Franks. Fax: +44 (0) 114 222 0002; e-mail:


Maximum stomatal conductance to water vapour and CO2 (gwmax, gcmax, respectively), which are set at the time of leaf maturity, are determined predominantly by stomatal size (S) and density (D). In theory, many combinations of S and D yield the same gwmax and gcmax, so there is no inherent correlation between S and D, or between S, D and maximum stomatal conductance. However, using basic equations for gas diffusion through stomata of different sizes, we show that a negative correlation between S and D offers several advantages, including plasticity in gwmax and gcmax with minimal change in epidermal area allocation to stomata. Examination of the relationship between S and D in Eucalyptus globulus seedlings and coppice shoots growing in the field under high and low rainfall revealed a strong negative relationship between S and D, whereby S decreased with increasing D according to a negative power function. The results provide evidence that plasticity in maximum stomatal conductance may be constrained by a negative S versus D relationship, with higher maximum stomatal conductance characterized by smaller S and higher D, and a tendency to minimize change in epidermal space allocation to stomata as S and D vary.


At their widest apertures, the stomatal pores on the surface of leaves determine the maximum leaf diffusive conductance (stomatal conductance) to CO2 and water vapour (gwmax and gcmax, respectively). Typical values for stomatal conductance span over two orders of magnitude across the full diversity of vascular plants (Körner, Scheel & Bauer 1979; Hetherington & Woodward 2003), and this tends to correlate with photosynthetic capacity (Field & Mooney 1986; Schulze et al. 1994). Within any particular plant group, the range of stomatal conductance is usually more restricted (Franks & Farquhar 1999; Wullschleger, Tschaplinski & Norby 2002; Long et al. 2004; Franks 2006), but in each case gwmax and gcmax are ultimately determined by the size (S, taken here as length L by width W) and density (number per unit area, D) of stomata in the epidermis. Despite widely documented variation in S and D across species (Hetherington & Woodward 2003), the constraints on the relationship between S and D are not well understood.

Short-term adjustment of stomatal aperture in response to fluctuations in external stimuli, such as light, CO2 or humidity, is achieved via an elaborate control mechanism involving ion channels in the guard cell membranes (Schroeder et al. 2001), as well as leaf photosynthetic and hydraulic signals (Farquhar, Dubbe & Raschke 1978; Buckley & Mott 2000; Meinzer 2002; Franks 2004; Mott, Sibbernsen & Shope 2008). By regulating stomatal aperture and hence the rate of transpirational water loss, plants can attempt to minimize their water deficit. Improvements in this regulatory mechanism have helped plants expand into a wide range of periodically water-limited habitats (Edwards, Kerp & Hass 1998; Raven 2002; Franks & Farquhar 2007). However, even with this complex stomatal aperture control mechanism, the large variations in S and D suggest that maximum stomatal conductance is tuned to local conditions. The advantage of this over simply producing a combination of S and D for highest gwmax, and adjusting aperture accordingly, is unclear.

There is considerable evidence for both short-term and long-term adjustments of S and D in response to environmental conditions. Salisbury (1928) identified higher D in sun leaves and xerophytes, as well as plants grown under low humidity or water limitation, a pattern now widely observed (Gindel 1969; Quarrie & Jones 1977; Clifford et al. 1995). Sensitivity of D to CO2 is also well documented, with growth chamber experiments and measurements over long-term CO2 changes using fossil stomata confirming a general decrease in D with elevated atmospheric CO2 (Woodward 1987; Beerling et al. 1993). Although S and D are potentially equally important in determining gwmax and gcmax, the focus to date has been overwhelmingly on environmentally induced variation in D. This has been enhanced by use of D (or its related measure, stomatal index) as a proxy for reconstruction of atmospheric CO2 concentration (Beerling, Birks & Woodward 1995; Royer et al. 2001). Stomatal index, introduced by Salisbury (1928), is defined as 100D/(DE + D), where DE is the number of epidermal cells per unit area. Stomatal size S is an important trait because it uniquely defines the upper limit of D. Stomatal size and density also determine total epidermal area allocated to stomata, which has implications for overall function of the epidermis, including how much area is available for other structures such as trichomes. A similar stomatal size measure, L2, when multiplied by D, was found to correlate with leaf hydraulic conductance (Sack et al. 2003), implying coordination with whole-plant attributes.

When measures of change in D and S are reported together, there is often a negative relationship between the two. That is, for different kinds of stimuli, including variation in water availability and atmospheric CO2 concentration, an increase in D is often accompanied by a decrease in S, and vice versa (Gindel 1969; Franks & Farquhar 2001; Uprety et al. 2002). Hetherington & Woodward (2003) proposed that small stomata will have shorter response times, and that this, in combination with their usually high densities, allows the leaf to attain high stomatal conductance rapidly under favourable conditions, but then to rapidly reduce conductance when evaporative conditions are unfavourable. While this would be an advantage over larger, slower stomata, it is unclear why a plant should revert back to larger, slower stomata under less demanding conditions, rather than simply adjust D and keep S small. Furthermore, gwmax and gcmax can be altered by various changes in the combination of S and D, so it is unclear if plasticity in stomatal conductance is necessarily constrained by a negative relationship between S and D.


Maximum stomatal conductance to water vapour (gwmax, in moles per metre squared per second; mol m−2 s−1) may be estimated using (Franks & Farquhar 2001):


where d is the diffusivity of water in air (m2 s−1), amax is the mean maximum stomatal pore area (m−2), v is the molar volume of air (m3 mol−1), l is the depth of the stomatal pore (m, approximated as W/2 for fully inflated guard cells (Franks & Farquhar (2007)) and π is the mathematical constant (typically approximated to 3.142). Maximum stomatal conductance to CO2, gcmax, is estimated as gwmax/1.6 (Farquhar & Sharkey 1982).

Equation 1 can be expressed in terms of stomatal guard cell length (L):


which can be re-arranged to calculate L from D and gwmax:


For Eqns 2 and 3, amax was taken as a fraction (α) of stomatal size (i.e. amax = αS). Although α varies across broad-ranging plant families (Franks & Farquhar 2007), a mid-range value of 0.12 was used for the demonstrations here. Equation 2 can also be given in terms of S:


Using Eqns 1–4, several basic relationships between S, D, gwmax, L and epidermal area allocation to stomata may be predicted. For a given gwmax (here, e.g. 1.0 mol m−2 s−1), smaller stomata allow considerably less epidermal area to be allocated to stomata, despite the requirement for more numerous stomata (increasing D) as S decreases (Fig. 1). Note also the negative power function relationship between D and S (Fig. 1a), and between D and L (Fig. 1b). Under these conditions (maintenance of gwmax as D and S vary), log transformation of the predicted D and S values yields a negative linear relationship between log D and log S (Fig. 2).

Figure 1.

For constant gwmax (here, 1.0 mol m−2 s−1), larger stomatal size S or length L requires a greater percentage leaf surface area allocation to stomata (solid lines in a and b, respectively). Under these conditions, the relationship between D and S (a), and between D and L (b) is in the form of a negative power function (dotted lines).

Figure 2.

Maintenance of constant gwmax (here, 1.0 mol m−2 s−1) with varying S and D results in a negative linear relationship between log10D and log10S (dotted line). However, achieving a given gwmax with larger S requires greater allocation of leaf surface area to stomata (solid line).

When total leaf surface (epidermal) area allocated to stomata is limited, then increasing gwmax is constrained by a negative relationship between stomatal density D and size S (Fig. 3). Under these conditions, increasing gwmax can only be achieved through decreasing S and increasing D. The other potential avenue for increasing gwmax, provided that the limit for epidermal surface allocation to stomata has not already been reached, is to increase S. However, this requires additional allocation of leaf surface area to stomata (Fig. 4). Even with this strategy, the maximum percent allocation of leaf surface area to stomata will eventually be reached, and further increase in gwmax will require a decrease in S and an increase in D (i.e. conformation to a negative linear relationship between log S and log D).

Figure 3.

Maintenance of constant epidermal area allocation to stomata (here, 25% of total epidermal area) results in a negative linear log–log relationship between stomatal density D and stomatal size S, and between gwmax and S.

Figure 4.

Provided stomata do not initially occupy the entire epidermal area; an increase in gwmax may be achieved by increasing S alone, or S and D, but eventually the limit of epidermal space allocation to stomata will be reached and further increase in gwmax can only be achieved by decreasing S and increasing D. Shown are plots of log10gwmax versus log10S for several constant percentages of leaf surface area allocation to stomata (25, 50, 100%). Increasing gwmax by increasing S requires a greater percent leaf surface area allocation to stomata.

The range in percent leaf surface area allocation to stomata across species is limited. In a comprehensive study of 90 woody and semiwoody species, it was found to be between about 2.8 and 33% (Cornelissen et al. 2003). Within a single species, it is much narrower, for example between 8 and 11% in tobacco grown under different light treatments (Rawson & Craven 1975), or between 0.84 and 1.32% for tulip poplar (Liriodendron tulipifera) at three different crown positions (McConathy 1983). If the allocation of epidermal space to stomata is limited, particularly to the extent noted above within species, then the relationship between S and D should generally be of the form depicted in Fig. 3. Under such conditions, the plastic range in gwmax would be constrained by a negative relationship between S and D.


Our objective was to examine the relationship between S, D and gwmax in a single species, Eucalyptus globulus. To examine trends across a range of values, we sampled leaves across a rainfall gradient, and from both seedling and coppice growth. The observed variability, which includes differences between individual leaves on the same plant, between leaves on different plants and between plants across environmental gradients, we refer to as ‘plasticity’, according to the definition in the review by Agrawal (2001). Our aim was to test for the following trends, based on the theory outlined above: (1) a negative power function relationship between S and D which, after log transformation, conforms to a negative linear relationship between log S and log D; (2) a positive correlation between gwmax (or gcmax) and D, and a negative correlation between gwmax and S; and (3) assuming a general positive correlation between stomatal conductance and leaf nitrogen concentration (Wong, Cowan & Farquhar 1985; Field & Mooney 1986; Schulze et al. 1994), a positive correlation between leaf nitrogen and D, and a negative correlation between leaf nitrogen and S. The findings are discussed in terms of the potential underlying spatial and energetic constraints, as well as broader functional implications.


Field sites and experimental design

All measurements were taken on second rotation, field-grown E. globulus at two plantations in south-western Australia: Avery's (34° 15′, 115° 31′) and Leighton's (34° 49′, 118° 33′). Both sites are characterized as Mediterranean, experiencing cool wet winters and hot dry summers. The average maximum and minimum air temperatures, 21 and 11 °C, respectively, are virtually the same for both sites (Fig. 5a). Likewise, the mean annual potential evaporation rates (E0) are similar at both sites, 1238 mm at Avery's and 1534 mm at Leighton's, and at both sites air temperature and vapour pressure deficit (VPD) can exceed 35 °C and 3.5 kPa, respectively, in summer (December to February). The sites differ markedly, however, in mean annual rainfall: 1033 mm at Avery's and 583 mm at Leighton's (Fig. 5b). This contrast is driven by different exposure to rain-bearing frontal systems originating from the Indian Ocean. At Avery's, the soil has a sandy-textured A horizon (0.5–1 m deep), over a clay B horizon, separated by a lateritic hard pan. At Leighton's, the A horizon (0.25–1 m deep) is of a sandy gravel texture, over a clay B horizon. Based on differences in mean annual rainfall, we henceforth refer to Avery's and Leighton's as ‘high rainfall’ and ‘low rainfall’ sites, respectively.

Figure 5.

Average monthly temperature (a) and rainfall (b) for the Avery's and Leighton's field sites (referred to in the text as high- and low-rainfall sites, respectively). The data are averages of records acquired between 1940 and 2008 (Australian Bureau of Meteorology 2008).

The first rotation was planted in 1996, and an experiment was established at each site in 1998 to explore the effects of nitrogen and thinning on leaf area development and response to drought. The plantations were originally planted at 1250 stems ha−1, and supplied with a basal fertilizer application at age 2 comprising phosphorus (100 kg ha−1 P), potassium (125 kg ha−1 K), magnesium (10 kg ha−1 MgSO4.7H2O), manganese (10 kg ha−1 MnSO4.H2O), zinc (10 kg ha−1 ZnSO4.7H2O) and copper (5 kg ha−1 CuSO4.5H2O), and an annual application of nitrogen (as urea) at 250 kg ha−1 N. First rotation trees were harvested (cut off at ground level with root systems left intact) in January 2006 (summer) at the high rainfall site, and June 2006 (winter) for the low rainfall site. Based on periodic measures of soil moisture throughout the profile, it was revealed that the trees had accessed the full depth of the B horizon, down to 6–10 m below the surface, and, at the time of harvest, had depleted the plant-available soil stored water down to this depth.

Following harvest of the first-rotation, six second-rotation experimental plots (three coppice plots and three seedling plots; 20 × 22 m) were established at each site by either: (1) replanting in July 2006 (winter wet season) with 7-month-old nursery-raised seedlings at a density of 1250 stems ha−1; or (2) allowing the first-rotation stumps to coppice. In the ‘seedling’ plots, coppicing was suppressed in the stumps remaining from the first-rotation harvest by painting them with glyphosate soon after harvesting. Stumps in the ‘coppice’ plots were allowed to naturally generate new shoots. A 50 g slow-release fertilizer tablet, comprising N, P, S, Mg, Cu and Zn, was applied to each individual soon after planting or coppice appearance. Nitrogen (as urea) was then applied annually at non-limiting rates in all treatments and sites (250 kg ha−1 year−1 N from year 2). An analytical assessment of leaves during the first rotation indicated that the remaining, potentially limiting, macro- and micronutrients were within the normal range for eucalypts (data omitted). All subsequent measurements were made on material obtained from the coppice and seedling measure plots in November and December 2007, corresponding to a period of optimal temperature for photosynthesis (Battaglia, Beadle & Loughhead 1996) and moderate VPD promoting stomatal opening (White, Beadle & Worledge 2000) in E. globulus.

Stomatal morphology

One leaf from five randomly selected plants per plot was collected at the high rainfall site, and one leaf from four randomly selected plants per plot was collected at the low rainfall site, yielding a total of 30 leaves (n = 15 from coppices; n = 15 from seedlings) for the high rainfall site, and 24 leaves (n = 12 from coppices; n = 12 from seedlings) for the low rainfall site. All leaves were classified as the ‘juvenile’ ontogenetic form. An epidermal impression was obtained halfway from the leaf tip to the base from the abaxial surface of each leaf by applying nail polish, allowing it to harden, and using clear cellophane tape to transfer the ‘impression’ to a microscope slide. Stomatal ratio was found to be less than 0.05, so our focus was on changes in abaxial stomatal properties.

Stomatal density [i.e. number of stomata per unit epidermal area (D, mm−2)] was calculated for each leaf as the mean of four fields of view at 400× magnification using a light microscope (Photomax LB; Olympus, Tokyo, Japan). Stomatal morphological parameters were measured as the mean of 20 stomatal complexes (guard cell pairs) for each leaf at 1000× magnification using the same light microscope. Stomatal morphological parameters were: guard cell length (L, µm), guard cell pair width (W, µm) and stomatal pore length (p, µm). To associate stomatal dimensions with stomatal conductance to water vapour (gw), which scales with pore area and density, we report stomatal size (S) as the product of L and W (µm2).

Leaf carbon isotope discrimination and nitrogen concentration

One newly expanded canopy leaf was collected from five randomly selected plants from each plot. After removing the midrib, the leaf tissue was oven dried at 70 °C, then finely ground with a ball mill. Exactly 0.050 g of ground material was then weighed, and 13C to 12C ratios and nitrogen elemental composition (% dry weight) were measured by means of a continuous flow mass spectrometer (model 20-20, IRMS; Europa, Crewe, UK). The carbon isotope composition of leaf tissue (δ13Cleaf) was calculated as:


where Rsample and Rstandard are the 13C/12C ratios of the leaf tissue and the V-PDB standard, respectively. δ13Cleaf was then converted to leaf carbon isotope discrimination (Δleaf) according to (Farquhar & Richards 1984):


where δ13Cair was taken as 7.8‰. On completion of analyses, the values for Δleaf and leaf nitrogen elemental composition were aggregated to yield an average for each plot.

Deriving time-integrated mean stomatal conductance

We sought to test for correlation, across sites, between gwmax and the time-integrated mean stomatal conductance at which leaves operated under natural conditions during shoot growth, inline image. A time-integrated measure of leaf intercellular CO2 mole fraction, inline image (in micromoles CO2 per mole air; µmol mol−1), was estimated from (Farquhar, O'Leary & Berry 1982):


where atmospheric CO2 mole fraction, ca, was taken as 370 µmol mol−1. Functions for leaf CO2 assimilation rate A versus leaf intercellular CO2 mole fraction ci were constructed using the model of Farquhar, von Caemmerer & Berry (1980) with the following constants (von Caemmerer 2000; Sharkey et al. 2007): (1) mitochondrial respiration rate (Rd) was 0.01Vcmax, in µmol m−2 s−1, where Vcmax is the maximum velocity of ribulose 1·5-bisphosphate carboxylase/oxygenase (Rubisco) for carboxylation (see below); (2) the CO2 compensation point in the absence of mitochondrial respiration (Γ*) was 37 µbar; (3) the Michaelis constant of Rubisco for carbon dioxide, KC, and oxygen, KO, was 27.3 Pa and 16.58 kPa, respectively; and (4) the partial pressure of oxygen was 21 kPa. This yielded an equation for estimating the time-integrated leaf CO2 assimilation rate, inline image (in micromoles per metre squared per second; µmol m−2 s−1):




and J (in µmol m−2 s−1) is the rate of carboxylation allowed by electron transport. Vcmax and J were estimated from the relationships for Vcmax versus leaf nitrogen for E. globulus (Warren 2004), and Jmax versus Vcmax (Wullschleger 1993), with J calculated according to von Caemmerer (2000). Time-integrated leaf conductance to CO2, inline image, was obtained from inline image, and time-integrated leaf conductance to water vapour, inline image, was then taken as inline image × 1.6 (Farquhar & Sharkey 1982). inline image comprises inline image in series with the time-integrated leaf boundary conductance, inline image, so inline image was subtracted from the series to give inline image (i.e. inline image. The mean wind speed for the study area was 3.2 m s−1 (Australian Bureau of Meteorology 2008), from which we estimated inline image to be approximately 1.0 mol m−2 s−1, based on data from several studies (Grace, Fasehun & Dixon 1980; Brenner & Jarvis 1995; Daudet et al. 1999).

Statistical analyses

The strength of bivariate relationships was quantified as correlation coefficients, and tests for significant differences among bivariate relationships were made by comparing the variation between regression models (Genstat version 11; VSN International, Hertforshire, UK). Non-linear least square curve-fitting procedures and graph plotting were carried out with data analysis and graphing software (OriginPro 8; OriginLab Corporation, Northampton, UK).


Stomatal size S and guard cell length L were negatively correlated with stomatal density D in both high-rainfall and low-rainfall populations (Fig. 6a,b). The negative, non-linear correlations between L and D (Fig. 6a), and between S and D (Fig. 6b) are well described by a log-normal model (L = 2075D0.90 + 9.67, r2 = 0.75 for high rainfall; L = 173D0.05 − 108, r2 = 0.55 for low rainfall, compared to S = 2.46 × 106D−1.64 + 189, r2 = 0.77 for high rainfall, S = 6.23 × 107D−2.62 + 348, r2 = 0.60 for low rainfall). Across all sites, the trend is consistent with the log-normal L versus D relationship reported in the survey by Hetherington & Woodward (2003) (dotted line in Fig. 6a). However, the relationship between S and D is best described as one in which S decreases with D in the form of a negative power function, as indicated by the strong linear relationship between log10S and log10D (Fig. 6c). Here, the distinctive characteristics of the high-rainfall and low-rainfall populations are more easily distinguished, with the slope and intercept of lines fitted by linear regression being substantially lower for the low-rainfall population (log10S = −1.04log10D + 5.17, r2 = 0.75, P < 0.001 for the high-rainfall site; log10S = −0.61log10D + 4.05, r2 = 0.61, P < 0.001 for the low-rainfall site). Assessment via linear regression with groups revealed that both the slope and y-intercept of the log–log plots were significantly different between the high- and low-rainfall sites (P = 0.034 for the slope, and P < 0.001 for the intercept).

Figure 6.

Guard cell length (L) and size (S) are negatively correlated with stomatal density (D) in Eucalyptus globulus. (a and b) Log-normal curves fitted to data from high-rainfall and low-rainfall sites. The slope and intercept of lines fitted by linear regression to log10S versus log10D are significantly different between high- and low-rainfall sites (c; see text for details).

The range of S and D differed between high- versus low-rainfall sites. Overall, there was a fivefold range in mean S and D (Fig. 6b), but the range was much broader for low-rainfall plants compared to high rainfall. Additionally, stomata in the seedlings tended to have smaller S and higher D compared to those in the leaves of coppices (t-test, P < 0.001 for S, and t-test, P < 0.001 for D). Within high- and low-rainfall populations, there is further separation in the data between coppice leaves and seedlings, with stomata in coppices being larger and at lower density compared to seedlings (Figs 6 & 7).

Figure 7.

The relationship between stomatal dimensions and percent leaf nitrogen, N. Stomatal size S is negatively correlated with N under high rainfall (a). Stomatal density D is positively correlated with N under both high and low rainfall (b). The slope and intercept of lines fitted by linear regression to D versus N are significantly different between high- and low-rainfall sites (b; see text for details).

A strong negative correlation was observed between S and percent leaf nitrogen N at high rainfall (Fig. 7a; S = −236N + 1240, r2 = 0.77, P = 0.02). A strong positive correlation between D and N was observed at both high- and low-rainfall sites (Fig. 7b; D = 85.9N − 22.8, r2 = 0.82, P = 0.013 for high rainfall; D = 213N − 244, r2 = 0.89, P < 0.010 for low rainfall). The slope and y-intercept of the relationships between D and N in high- versus low-rainfall sites were significantly different (P = 0.015 for the slope, and P = 0.009 for the intercept; linear regression with groups).

The mean maximum stomatal conductance (gwmax) was negatively correlated with S under low rainfall (Fig. 8a; S = −237gwmax + 837.25, r2 = 0.34, P < 0.010), and positively correlated with D under both high and low rainfall (Fig. 8b; D = 103gwmax + 21.1, r2 = 0.33, P < 0.001 for high rainfall, D = 202gwmax − 92.6, r2 = 0.82, P < 0.001 for low rainfall). Because of the differences in the range of S and D for high- and low-rainfall sites, the range of gwmax at the high-rainfall site (1.74–2.63 mol m−2 s−1) was smaller than at the low-rainfall site (0.71–2.40 mol m−2 s−1). The slope and y-intercept of the correlations between D and gwmax were significantly different between the high- and low-rainfall sites (P = 0.008 for the slope, and P < 0.001 for the y-intercept; linear regression with groups).

Figure 8.

The relationship between stomatal dimensions and maximum leaf conductance to water vapour, gwmax. Stomatal size S is negatively correlated with gwmax under high rainfall (a). Stomatal density D is positively correlated with gwmax under both high and low rainfall (b). The slope and intercept of lines fitted by linear regression to D versus gwmax are significantly different between high- and low-rainfall sites (b; see text for details).

Across all sites, gwmax increases in a saturating fashion with N (Fig. 9a; gwmax = 2.25 + 0.97e−(N−1.92)/0.58, r2 = 0.76; non-linear least square fit). High-rainfall plants distinctively occupy the upper part of this curve. Similarly, gwmax is positively correlated with leaf carbon isotope discrimination, Δleaf (Fig. 9b), with plants from high-rainfall plots showing highest Δleaf. These patterns apply similarly to the estimated mean operating stomatal conductance, inline image, with a non-linear correlation between gwmax and inline image across sites (Fig. 10; linear regression of log gwmax versus log inline image; y = 0.39 + 0.42x; r2 = 0.57, P = 0.004).

Figure 9.

Maximum stomatal conductance to water vapour (gwmax) increases in a saturating fashion with percent leaf nitrogen (N; a) and leaf carbon isotope discrimination (Δleaf; b). Plants from the high-rainfall sites had higher gwmax, N and Δleaf.

Figure 10.

Across all sites, maximum stomatal conductance to water vapour (gwmax), estimated from S and D measurements, is positively correlated with time-integrated mean stomatal conductance (inline image) estimated from leaf δ13C.

Bulk means for all of the leaf parameters are summarized in Table 1. These means reveal little about the interplay between S, D and gwmax, but allow a quick overall assessment of some basic differences between leaves under the different conditions. Consistent with the typical response to water deficit, leaves that grew under low rainfall exhibited significantly lower gwmax, inline image and Δleaf. They also had lower N. Between the two growth forms (seedling versus coppice), seedlings had significantly smaller S and higher D, as well as higher gwmax, inline image and N.

Table 1. Eucalyptus globulus leaf properties (mean ± SE)
ParameterHigh rainfallLow rainfallCoppiceSeedling
  1. Asterisks indicate significant difference between low rainfall and high rainfall, or seedling and coppice [two-way analysis of variance (anova), site by growth form, 0.05 level of significance, Tukey comparison of means].

  2. L, guard cell length; W, guard cell pair width; S, stomatal size (= L × W); amax, area of maximally open stoma assuming circular pore with diameter equal to pore length; gwmax, maximum stomatal conductance to water vapour, based on amax; inline image, estimate of time-integrated mean stomatal conductance to water vapour; N, percent leaf nitrogen; Δleaf, leaf carbon isotope discrimination.

L (µm)25.108 ± 0.175 (n = 600)22.991 ± 0.242* (n = 480)26.337 ± 0.212 (n = 540)21.996 ± 0.156* (n = 540)
W (µm)20.555 ± 0.138 (n = 600)19.299 ± 0.206* (n = 480)21.904 ± 0.170 (n = 540)18.089 ± 0.127* (n = 540)
S (µm2)522.683 ± 23.339 (n = 30)459.829 ± 39.043 (n = 24)586.931 ± 31.455 (n = 27)402.565 ± 17.482* (n = 27)
D (mm−2)238.853 ± 5.010 (n = 120)229.592 ± 10.945 (n = 96)192.533 ± 6.213 (n = 120)276.941 ± 7.352* (n = 96)
amax (µm2)165.291 ± 2.667 (n = 600)125.711 ± 2.866* (n = 480)177.633 ± 3.176 (n = 540)117.767 ± 1.815* (n = 540)
gwmax (mol m−2 s−1)1.660 ± 0.035 (n = 30)1.250 ± 0.076* (n = 24)1.372 ± 0.078 (n = 27)1.584 ± 0.048* (n = 27)
inline image (mol m−2 s−1)0.370 ± 0.018 (n = 30)0.192 ± 0.015* (n = 24)0.243 ± 0.020 (n = 27)0.334 ± 0.026* (n = 27)
n (%)3.056 ± 0.099 (n = 30)2.205 ± 0.088* (n = 24)2.370 ± 0.089 (n = 27)2.960 ± 0.130* (n = 27)
Δleaf (‰)21.325 ± 0.160 (n = 30)19.565 ± 0.293* (n = 24)20.284 ± 0.298 (n = 27)20.781 ± 0.258 (n = 27)


The relationship between stomatal size S and density D in E. globulus followed the form of a negative power function, as confirmed by linear regression of log S on log D (Fig. 6c). The slopes of the regressions were shallower than that required for maintenance of constant gwmax, indicating that the range in S versus D was associated with plasticity in gwmax. The slope of the linear log–log regressions also differed between plants from high-rainfall versus low-rainfall zones, but data for seedling and coppice plants fell on the same line for each respective rainfall zone. This suggests that the constraints on plasticity of gwmax imposed by environmental moisture are different from those imposed internally by the contrasting whole-plant developmental conditions of seedling versus coppice growth.

The observed negative relationship between S and D forms the basis of a negative correlation between gwmax and S, and a positive correlation between gwmax and D (Fig. 8). The overall pattern is a tendency for higher gwmax to be achieved through a decrease in S and an increase in D (i.e. smaller, more numerous stomata). As predicted on the basis of known correlations between plant gas exchange capacity and leaf nitrogen concentration N (see Introduction), the observed correlations between gwmax, D and S were accompanied by positive correlations between N and D, and negative correlations between N and S (Fig. 7). However, as with the overall relationship between S and D, the difference between high- versus low-rainfall plants in terms of correlations between gwmax, S and D, as well as between N, S and D, suggests strong coordination of gwmax in developing leaves of E. globulus by environmental moisture. Leaves that matured under low rainfall had smaller stomata at any given D (Fig. 6), consistent with a down-regulation of gwmax (Fig. 8a), and a shift towards higher water-use efficiency (WUE), indicated by lower Δleaf (Fig. 9b).

With regard to coppice versus seedling leaves, it is more commonly observed that coppice leaves have higher stomatal conductance and lower WUE than leaves of uncut stems (Kruger & Reich 1993; Wildy, Pate & Sefcik 2004), with the explanation being that coppices are attached to a disproportionately large root system that keeps them well supplied with water. However, when E. globulus trees are grown repeatedly at the same site in a plantation setting, there is potential to deplete deep-soil moisture reserves (Drake et al. 2009). Thus, the deep-root system of second-rotation coppices may be positioned in drier soil relative to shallow-rooted, second-rotation seedlings that can derive moisture from recent rainfall events. In our study, coppice leaves had lower gwmax than seedlings in both high-rainfall and low-rainfall sites, which is consistent with the likelihood that coppices regenerated under conditions of greater water deficit.

The strong correlation across sites between modelled gwmax (obtained from S and D measurements) and inline image (obtained from leaf δ13C and leaf nitrogen) (Fig. 10) demonstrates that patterns of plasticity in S and D translate directly to differences in the operational stomatal conductance for plants growing in different conditions. Thus, leaves built with S and D combinations that yield higher gwmax appear to operate with higher stomatal conductance. The mean operating stomatal conductance, represented by the proxy inline image, is considerably less than modelled gwmax for each site (Fig. 10), due at least in part to the tendency of reduced humidity, wind, soil water deficit and even ambient CO2 to reduce stomatal apertures well below the maximally opened state represented by gwmax.

Our results further demonstrate that it is not only environmental conditions that drive the plasticity of S, D, and hence gwmax, and inline image in developing leaves. Like many Eucalyptus species, E. globulus produces juvenile-form leaves in both seedling and coppice growth (Boland et al. 2006). However, although in this study the new leaves in seedlings and coppices were developing under similar environmental conditions at either high-rainfall or low-rainfall sites, stomatal development was distinctively different between the two. It is therefore likely that S and D are determined not only by conditions in the vicinity of the developing leaf, but also remote from it, in this case the adult rootstock. This is consistent with the observation by Lake et al. (2001) that adult leaves can signal adaptive changes in the morphology of stomata of new leaves independent of the environment experienced by the new leaves.

The pervasive negative correlation between S and D might be explained by physical and energetic constraints. The physical constraints relate to the simple spatial requirements for embedding stomata of sufficient size and density into the leaf epidermis for the optimal gwmax, while satisfying a given stomata to pavement cell ratio (Figs 3 & 4). The energetic constraints relate to the return, in terms of gcmax and dynamic stomatal properties, and therefore photosynthetic productivity and WUE, for a given investment in stomatal infrastructure.

As shown in Fig. 2, various combinations of S and D can yield the same gwmax. In Fig. 11a are loci (light lines) of log10S versus log10D for several values of gwmax. As long as the average stomatal size for a given density falls on the relevant gwmax locus, there can be considerable natural variation in S and D while maintaining constant gwmax. Alternatively, if epidermal space allocation to stomata is constrained within narrow limits or maintained constant, variation in gwmax also results in a negative relationship between S and D (light lines in Fig. 11b), except with a shallower slope compared to the condition of constant gwmax. The thick, dark and dashed lines in Fig. 11a, representing the data in Fig. 6c, cut across constant gwmax loci, indicating plasticity in gwmax, with their negative slope showing that higher gwmax in E. globulus is achieved with smaller S and higher D. Under high rainfall, plasticity in gwmax is achieved with relatively constant epidermal area allocation to stomata (a little over 12%; thick, dark line in Fig. 11b), but under low rainfall reduced gwmax is achieved with reduced epidermal area allocation to stomata (dropping to below an estimated 6%; thick, dashed line in Fig. 11b), consistent with a more water-conservative leaf structure.

Figure 11.

The relationship between S and D for (a) different constant values of gwmax and (b) different constant values of percent epidermal area allocated to stomata, represented as a series of straight lines on a plot of log S versus log D[here, shown for (a) gwmax = 0.5, 1.0 and 2.0 mol m−2 s−1, and (b) area = 3, 6 and 12% of epidermis; see Introduction for underlying equations]. In theory, different leaves can exhibit a range of S and D combinations, but retain the same gwmax (a; see also Figs 1 & 2) or vary gwmax and retain the same % area allocation to stomata (b; see also Figs 3 & 4), or any variation between these two strategies. Superimposed on both panels are lines describing the correlations in Fig. 6c for high- and low-rainfall sites. High-rainfall plants tended to maintain constant percent epidermal area allocation to stomata while varying gwmax, whereas low-rainfall plants extended their plastic range of gwmax into lower values while reducing percent area allocation to stomata. In both high- and low-rainfall plants, higher gwmax was associated with smaller S and higher D.

The low-rainfall leaves span a greater range of gwmax than high rainfall, with a similar upper limit, but extending down to much lower gwmax values (Fig. 8). In both cases, the correlations reveal that leaves adopt the strategy of producing higher numbers of smaller stomata to increase gwmax, or vice versa, lower numbers of larger stomata to decrease gwmax. In theory, gwmax can be increased by producing leaves with higher D alone, or with higher D and larger S, but there is no evidence for this in our results. The most likely reason is competition for space on the epidermis between stomata and non-stomatal epidermal cells, including trichomes and glands. Increasing gwmax through a decrease in S and an increase in D allows minimal change in the area of epidermis occupied by stomata. If, however, gwmax is increased by increasing D and holding S constant, or by increasing both S and D, the area occupied by stomata must increase dramatically, impacting on space that might be needed by other epidermal cell structures. Thus, it appears that changes in S, D and hence gwmax are constrained by a requirement for minimal change in allocation of space on the leaf surface to stomata.

If the potentially improved dynamic attributes of smaller somata (Hetherington & Woodward 2003) are a factor that drives the negative S versus D relationship, then this should be manifested as improved long-term WUE. Our measurements of Δleaf, a surrogate for time-integrated WUE (Farquhar, Ehleringer & Hubick 1989), are consistent with this. At similar gwmax, stomata in the low-rainfall leaves were smaller and D was larger (Fig. 8), and these leaves also had lower Δleaf and therefore higher WUE. This could represent a further selective pressure in favour of the negative S versus D relationship.

If, for example, all guard cell membrane properties were equal, then smaller guard cells might be able to change osmotic and turgor pressure more quickly, leading to more responsive stomata. This could be attributed to the improved surface-area-to-volume ratio of smaller stomata. However, for the same gwmax, more numerous stomata are required if they are to be smaller (Figs 1 & 2), which raises the question of whether there is added metabolic cost for running the additional duplicates of complete stomatal metabolic machinery, even though individual and total stomatal cell volume per leaf is lower. There is limited quantitative information on this topic, but higher guard cell respiration rates have been reported in high conductance lines of Pima cotton (Srivastava, Lu & Zeiger 1995), suggesting perhaps higher metabolic cost for higher gwmax. This would not be a disadvantage if paid for with higher rates of CO2 assimilation, and might explain reversion back to fewer larger stomata with lower gwmax, particularly if driven by a reduction in a resource directly affecting photosynthesis, such as low light or nitrogen availability.


While many combinations of S and D can yield the same gwmax, our results show that when gwmax in new leaves is altered within its plastic range because of environmental or internal factors (rainfall, nitrogen, crown loss), it is done so via a negative relationship between S and D, in which S decreases with increasing D in the form of a negative power function. This is consistent with a general relationship of the same form reported by Franks & Beerling (2009) across multiple species and geological time-scales, suggesting a fundamental inter-relationship between S and D that governs short-term (plastic) and long-term (evolutionary) adaptations of gwmax (and gcmax) to environment. One explanation for this relationship is improved economy of space allocation on the leaf surface to stomata. Despite the potential advantages of larger numbers of smaller stomata (higher gwmax and improved dynamic performance), the broad-based negative relationship between S and D implies that there is a cost associated with high numbers of small stomata, and reversion back to smaller numbers of larger stomata is a better strategy for conditions in which lower gwmax is sufficient. Further work is necessary to quantify the added cost of increased numbers of smaller stomata, and to determine whether it is purely energetic or of a broader functional nature.


We are grateful for the technical assistance provided by Jessie Rutter, and to the Albany Plantation Forest Company and Hansol PI for site access. We acknowledge support from the Co-operative Research Centre for Forestry and the Australian Research Council.