Comparing model predictions and experimental data for the response of stomatal conductance and guard cell turgor to manipulations of cuticular conductance, leaf-to-air vapour pressure difference and temperature: feedback mechanisms are able to account for all observations



    Corresponding author
    1. Institute for Water and Environmental Resource Management, Department of Environmental Sciences, University of Technology Sydney, PO Box 123, NSW 2007, Australia and
      D. Eamus. Fax: 61 2 95144079; e-mail:
    Search for more papers by this author

    1. Institute for Water and Environmental Resource Management, Department of Environmental Sciences, University of Technology Sydney, PO Box 123, NSW 2007, Australia and
    Search for more papers by this author

    1. Institute for Water and Environmental Resource Management, Department of Environmental Sciences, University of Technology Sydney, PO Box 123, NSW 2007, Australia and
    Search for more papers by this author

    1. Charles Darwin University, Darwin, NT 0909, Australia
    Search for more papers by this author

    1. Institute for Water and Environmental Resource Management, Department of Environmental Sciences, University of Technology Sydney, PO Box 123, NSW 2007, Australia and
    Search for more papers by this author

D. Eamus. Fax: 61 2 95144079; e-mail:


Stomata respond to increasing leaf-to-air vapour pressure difference (LAVPD) (D) by closing. The mechanism by which this occurs is debated. A role for feedback and peristomatal transpiration has been proposed. In this paper, we apply a recent mechanistic model of stomatal behaviour, and compare model and experimental data for the influence of increasing D on stomatal conductance.

We manipulated cuticular conductance (gc) by three independent methods. First, we increased gc by using a solvent mixture applied to both leaf surfaces prior to determining stomatal responses to D; second, we increased gc by increasing leaf temperature at constant D; and third, we coated a small area of leaf with a light oil to decrease gc. In all three experiments, experimental data and model outputs showed very close agreement.

We conclude, from the close agreement between model and experimental data and the fact that manipulations of gc, and hence cuticular transpiration, influenced gs in ways consistent with a feedback mechanism, that feedback is central in determining stomatal responses to D.


The behaviour of stomatal conductance (gs) is central to the maintenance of plant water status (Buckley & Mott 2002; Buckley 2005; Roelfsema & Hedrich 2005) because it is the diffusion of water vapour at the leaf epidermis, through stomata and across the cuticle, which determines the rate of water loss from plants. At a constant leaf-to-air vapour pressure difference (LAVPD) (D), increased stomatal or cuticular conductance increases transpiration rate and decreased conductance will decrease transpiration. Leaves respond to varying environmental conditions by varying gs to maintain a favourable water balance and prevent xylem embolism (Sperry & Pockman 1993; Hubbard et al. 2001; Dewar 2002; Brodribb & Holbrook 2004).

Increasing D leads (after a transitory increase) to decreased gs (Davies 1986; Oren et al. 1999; Eamus & Prior 2001; Buckley 2005), while decreasing D leads to increased gs (Nonami, Schulze & Ziegler 1990; Schulze 1993). Despite several decades of research, there is no accepted mechanism to explain how guard cells detect D (Grantz 1990; Franks, Cowan & Farquhar 1997; Franks 2004; Powles et al. 2006), but it is most likely that stomata respond to changes in transpiration rate (E), rather than exhibiting a direct response to D (Mott & Parkhurst 1991; Monteith 1995). However, D is routinely used experimentally to influence gs (Schulze 1993; Oren et al. 1999).

Stomatal responses to changes in E (driven by changes in D) show three distinct phases (Monteith 1995). At low values of D, transpiration increases as D increases (phase C) and the limitation on E exerted by stomata is minimal. Stomata begin to increasingly limit E at intermediate values of D, and E remains constant or increases only minimally as D increases (phase A). At larger values of D, increases in D cause large declines in gs, leading to reduced E (phase B). During drought conditions, phases C and A may be eliminated so that E decreases with increasing D (Thomas & Eamus 1999).

Several mechanisms for stomatal response to D (or E) have been proposed (Buckley 2005; Powles et al. 2006). Monteith (1995) concluded that stomatal closure associated with increased D is caused by an increase in the rate of transpiration from the whole leaf and concomitant reductions of leaf water status. This finding followed the work of Mott & Parkhurst (1991), which showed that when D increased but transpiration rate remained constant (by using Helox), gs did not change, indicating the importance of leaf water status. However, varying the leaf water potential independently of transpiration rate did not affect the response of gs to D (Bunce 1996), so the importance of leaf water status appears variable, perhaps because the role of whole-plant hydraulics or root-to-shoot communication in stomatal sensitivity and minimum leaf water potential needs to be considered (Addington et al. 2004; Franks 2004).

Mott & Parkhurst (1991) could not distinguish between transpiration through stomata and transpiration across the cuticle. An effect of cuticular and peristomatal (across the cuticle of the guard cells; Maier-Maercker 1979), transpiration on gs has been proposed (Grantz 1990; Bunce 1996; Eamus & Shanahan 2002) as a mechanism for the reduction in transpiration with increasing D. Kerstiens (1996, 1997) highlights the importance of cuticular conductance in determining stomatal responses to D, while Mott & Franks (2001) highlight the role of the turgor of subsidiary/epidermal cells relative to that of guard cells (the antagonism ratio, Raschke 1979; the mechanical advantage, Franks 2004) in explaining stomatal responses to D. Mott & Franks (2001) showed transient opening of stomata in response to a sudden increase in D over a small patch of leaf. Simultaneously, local epidermal cell turgor declined and it was concluded that increased stomatal transpiration reduced epidermal cell turgor and the mechanical advantage (or antagonism ratio) was such that stomatal opening was increased as a result of decreased epidermal turgor, relative to guard cell turgor. However, the possibility that it was increased cuticular transpiration, rather than stomatal transpiration, that caused epidermal cell turgor to decline, was not discussed by Mott & Franks (2001). Farquhar (1978) showed that a response of gs to cuticular transpiration could explain stomatal regulation of transpiration that was independent of mesophyll or xylem leaf water potential and could explain declining transpiration with increasing D (the so-called feed-forward response of stomata; Raschke 1979; Grantz 1990; Franks et al. 1997). A role for cuticular transpiration in the mechanism linking increased D to a stomatal response, therefore, remains controversial. Meinzer (1982) showed that increased cuticular conductance (gc) increased stomatal sensitivity to D, but Frensch & Schulze (1988) found no change in epidermal cell turgor or E in response to changes in D, but these experiments were conducted with closed stomata. Kerstiens (1997) concluded that increased gc did not generally increase the sensitivity of stomata to D, and that there was a minimal role for peristomatal transpiration in stomatal responses to D. However, he does suggest that increased gc (and hence, cuticular transpiration) reduces epidermal cell turgor, and this reduces the back pressure exerted on guard cells, which may explain why increased gc did not increase stomatal sensitivity to D (Kerstiens 1997). Clearly, the role of gc in stomatal responses to D, and the role of epidermal (or more properly, subsidiary cell) turgor, remain unclear. In this paper, we provide experimental and modelling data supporting the view that gc plays a central role in the feedback between D and gs.

Stomatal responses to D have been modelled on the basis of changes in gradients of water potential within the epidermis and xylem (Dewar 2002; Eamus & Shanahan 2002) or on the basis of a metabolic response within guard cells following a change in the local (epidermis plus guard cells) water status arising from a change in D (Buckley, Mott & Farquhar 2003). Both of these models were able to replicate the three-phase response of gs to D, as reviewed by Monteith (1995) and as shown experimentally (Thomas & Eamus 1999). The Eamus & Shanahan (2002) model explains the stomatal response to E, in terms of the supply of water to the guard cell from epidermal cells. More importantly, it predicts that changes in gs, in response to changes in D or temperature, can be explained by the influence of these factors on gc, and that changes in the ratio of total transpiration rate to cuticular transpiration rate influence gs. We apply this model, with one minor modification, to address the following questions: what are the modelled and experimental responses of gs to changes in D; how does a change in gc influence the modelled and experimental responses of gs to changes in D; what are the modelled and experimental responses of gs to changes in temperature under constant D; what are the modelled and experimental responses of the ratio of stomatal transpiration rate to cuticular transpiration rate to changes in D, gc and temperature? The only modification to the Eamus & Shanahan (2002) model was to incorporate the strong temperature dependence of cuticular conductance (Schonherr, Eckl & Gruler 1979; Kerstiens 1996).


Plant growth conditions

Four experiments were conducted. In the first and second, the response of gs to an increased gc and increasing leaf temperature at constant D were examined using leaves of Eucalyptus haemastoma (Sm.), a native tree species common to New South Wales of Australia. In the third experiment, the response of stomatal aperture to a decrease in gc was examined using three species, Commelina communis, Commelina cyanea and Vicia faba. In the fourth experiment, the response of guard cell turgor to changes in water vapour pressure above a leaf was examined using Rhoeo discolor. These later four species were chosen because they have large stomates and have been extensively used in studies of stomatal behaviour in the past.

Eucalyptus haemastoma (Sm.) seedlings were purchased from local nurseries as 4- to 7-month-old container plants (210 mL pots) and kept in a rooftop glasshouse for 1–5 months before experimentation began. Maximum photosynthetic photon flux density (PPFD), temperature and RH experienced in spring (October 2003) were 1015 µmol m−2 s−1, 29 °C and 91% RH, respectively, and in Autumn (May) were 630 µmol m−2 s−1, 26.2 °C and 95.5% RH, respectively. All plants were repotted into 500 mL pots containing a 1:1:1 mix of potting mix (Hortico, Orica Plc, Sydney, Australia), perlite and sand. After repotting, all plants were kept well-watered and under a shade cloth to reduce transpiration, for at least 2 weeks, to allow recovery from root disturbance before being returned to normal glasshouse conditions. All plants were kept well-watered by receiving irrigation to drip point twice daily. All other species were grown from seed in the same glasshouse as described earlier, in the same size pots with the same potting mix.

Experiment 1: stomatal responses to D at constant temperature and two values of gc

The response of gs of E. haemastoma leaves to increasing and decreasing D was measured using a laboratory-based open-gas exchange system. The system consisted of three aluminium leaf cuvettes (internal dimensions 150 × 110 × 75 mm) with glass tops. The air within each cuvette was stirred with a small fan to minimize leaf boundary layer resistance. Leaf temperatures were regulated by circulating cooled water through the floor of each cuvette. Leaf temperatures were measured with K-type thermocouples (one per leaf) adpressed to the adaxial surface. Illumination was provided by 400 W metal halide lamps (one per cuvette) overhanging each cuvette. PPFD was measured inside each cuvette before each experiment with a Li-Cor quantum sensor and data logger (LI-1400; Li-Cor, Lincoln, NE, USA). Cuvette airflow rates were regulated using mass flow controller (Brooks Smart Series 5850s; Brooks Instruments, Hatfield, PA, USA) placed upstream of each cuvette. A cooled-mirror dew point hygrometer (Hygro-M1; General Eastern Instruments, Watertown, MA, USA) measured the dew point temperature of the air entering and exiting each cuvette.

LAVPD (D) was manipulated by varying the dew point of the inlet air. Before entering the cuvettes, the inlet airstream was bubbled through a temperature-controlled water bath. The inlet dew point was altered by changing the temperature of the water bath. Leaf area was measured with a leaf area meter (Li-Cor 3000A).

Measurements of stomatal responses to changing D were collected from one leaf from each of three replicate plants. Plants were moved from glasshouse to laboratory the afternoon before experimentation began. Replicates were chosen on the basis of uniformity of height (35–40 cm) and the absence of any visual signs of damage, nutrient deficiency or disease. For each replicate, a single, young, but fully expanded leaf (usually the third or fourth leaf from the shoot apex) was placed within a cuvette. Light intensity at the leaf surface was increased at 1000 h to a maximum of 1000 µmol m−2 s−1, over a period of 90 min, by removing neutral density filters (preliminary experiments had shown this PPFD to be saturating). Leaf gas exchange was measured over a period of 5–6 h (between 1000 and 1600 h) as D was varied between 1.0 and 4.0 kPa in approximately 0.3 kPa steps while maintaining leaf temperature at 30 °C (±1 °C). Leaf gas exchange was recorded 60 min after D was varied. During the experiment, plants were kept well-watered by maintaining a small reservoir of water under each pot.

The response of gs to D was determined on control leaves, which received no pretreatment and on leaves which were pretreated with a hexane–ethanol mix to increase gc. This method of manipulating gc is now described.

In order to test the influence of gc on the stomatal response to increased D, young but fully expanded leaves of E. haemastoma (Sm.) were wiped twice on both sides of the leaf with tissue paper that had been soaked in a mix of hexane and ethanol, in a 1:2 ratio and then squeezed so that it would not drip (Kerstiens 1997). Prior to wiping with the tissue, leaves were placed in darkness for 40–60 min to minimize the potential for uptake of the solvents through stomatal pores (Kerstiens 1997). This application of hexane/ethanol temporarily increases cuticular conductance (Kerstiens 1997). Leaves were then exposed to an increase in D (as described earlier) 24–30 h after solvent application. The same leaves were subjected to an increase in D 48 h prior to the application of solvent to establish control response curves. Preliminary studies have established this age difference between control and solvent-treated leaves had no effect on stomatal behaviour. Leaves of E. haemastoma are long lived (>12 months).

In order to quantify the effect of the solvent treatment on cuticular conductance, solvent-treated and control leaves were allowed to transpire while attached to the plant, at a constant D of approximately 2.0 kPa for 30–60 min. Leaves were then excised and subjected to darkness by covering the cuvettes in aluminium foil (excision and darkness promote stomatal closure). Leaf conductance was calculated by dividing the rate of water loss from the leaf by the LAVPD (expressed in mole fraction of water vapour) repeatedly over a 320 min period. A plot of leaf conductance against time shows a rapid rate of decline in leaf conductance as the stomata close. Once stomata are closed (typically 30–40 min), the rate of change of leaf conductance is extremely small because the cuticular conductance is essentially constant. Cuticular conductance was estimated by extrapolating the linear final portion of the curve to the y intercept. The effect of the solvent treatment on cuticular conductance was then determined by comparison of the cuticular conductance of control and solvent-treated leaves.

Experiment 2: the response of gs, gc and Ci to leaf temperature at constant D

The response of gs and transpiration rate to increasing temperature at constant D was obtained by increasing leaf temperature over the range 18–38 °C (in steps of 2–3 °C), while maintaining D constant at 2.1 kPa. Leaf temperature was gradually increased (over a 10 min period) by increasing the temperature of the temperature regulation bath, and D was held at 2.1 kPa by simultaneously increasing the water vapour pressure of the inlet airstream (by raising the temperature of the vapour pressure control bath). Leaves were allowed 30–40 min to adjust to each step increase in temperature before stomatal and gas exchange responses were recorded. Light flux density was maintained at 1000 µmol m−2 s−1 throughout. Ci was calculated for each leaf at each temperature from the relationship between assimilation rate and gs. At the end of this process, the leaves were excised and darkened within the cuvette, and leaf temperature reduced to the initial temperature (approximately 19 °C). Stomatal closure occurred within 40 min, and the experiment was repeated to establish the response of cuticular conductance to increasing leaf temperature at constant D. There was no evidence of stomata remaining open after 50 min in darkness when leaves were examined with a microscope.

Experiment 3: the influence of a decreased cuticular conductance on stomatal aperture

For three species (C. communis, C. cyanea and V. faba), three replicate intact leaves were gently clamped abaxial side up, on the stage of a microscope, at two light intensities [high ca. 750 µmol m−2 s−1 photosynthetically active radiation (PAR) and moderate ca. 500 µmol m−2 s−1 PAR]. The microscope was fitted with a video camera and video recorder. Stomatal apertures of typically 15–20 guard cell pairs of each leaf were recorded before the application of a small quantity (ca. 0.1 µL) of light oil (microscope lens immersion oil). Stomatal aperture was recorded before and 15 min after the addition of the oil.

Experiment 4: guard cell turgor potential as a function of D

A micropressure probe (Eamus & Wilson 1984) was used to measure the turgor of a guard cell from three stomatal complexes of one leaf from each of three replicate R. discolor plants. The vapour pressure deficit (VPD) of air above an area of leaf of approximately 2 cm2 was controlled by passing air of known water VPD (1.5, 2.5, 3.5 kPa) over the surface of the leaf from the needles of six syringes held in parallel and angled towards the leaf. Illumination of the leaf surface was from a cool light fibre optic system. Leaf temperature was approximately 30 °C and was recorded with a fine thermocouple adpressed to the surface of the leaf. Leaves were maintained at each level of LAVPD for 20 min prior to measurement of turgor pressure.

Statistical analysis

Standard linear or polynomial functions were fitted to observed responses of gs and transpiration to increasing D. Differences in cuticular conductance, between solvent-treated and control leaves, were detected using analysis of covariance (ancova). Differences in stomatal aperture resulting from the addition of oil to the leaf surface, and differences in turgor pressure arising from exposure to different D above the leaf surface were compared with a paired t-test. All data were tested for and satisfied the assumptions of normality and homogeneity of variance. All differences were evaluated at a significance level of 0.05.


Stomatal conductance and transpiration: intact, untreated leaves

As D increased over the range 1.4–3.0 kPa, gs showed a curvilinear decline from a mean of 264 mmol m−2 s−1 (at 1.4 kPa) to a final value of 210 mmol m−2 s−1 (at 3.0 kPa; Fig. 1a). Transpiration rate showed a curvilinear increase, from a mean minimum of 3.6 mmol m−2 s−1 (1.4 kPa) to a mean of 6.13 mmol m−2 s−1 (3.0 kPa) (Fig. 1b). The response of gs to increasing and decreasing D was examined and found not to differ (Fig. 1a).

Figure 1.

(a) Stomatal conductance increases to a maximum and then declines as D increases, for leaves of Eucalyptus haemastoma; (b) transpiration increases to a maximum value and then shows a small decline as D increases. The linear and quadratic terms for the polynomial regressions under increasing leaf-to-air vapour pressure difference (LAVPD) (both for gs and transpiration) were significant at P < 0.0001. There was no difference between the curves for increasing and decreasing D.

The effect of hexane–ethanol on cuticular conductance

Before excision and simultaneous darkening treatment, leaf conductance remained constant (data not shown) in both control and hexane–ethanol-treated leaves. Upon excision from the stem and subsequent darkening, leaf conductance of both control and solvent-treated leaves declined quickly to about 5% of the initial mean value after 30–40 min (data not shown). Data for the period 80–300 min after excision were used to calculate gc. Cuticular conductance values of solvent-treated and control leaves were 8.3 and 2.50 mmol m−2 s−1, respectively (P < 0.01). The solvent treatment caused an increase in cuticular conductance of ca. 250%. We did not test for the effect of wiping the leaves with a tissue soaked in water, and therefore, we cannot discount the possibility that the cause of the observed increase in gc was the act of wiping rather than the presence of the solvent.

Cuticular conductance and the stomatal response to D

There was a qualitative difference in the response of stomatal conductance to an increase in D between control and solvent-treated leaves (Fig. 2). As D increased, gs declined linearly for treated leaves, whereas the control curve was convex (Fig. 2). Thus, for treated leaves, the response of gs to increasing D was constant over the entire range of D, but for control leaves the response of gs to D increased with increasing D. Furthermore, the gs of leaves treated with solvent was consistently smaller than that of control leaves, for all values of D (Fig. 2).

Figure 2.

The response of stomatal conductance of control leaves to increasing D differs from the response observed for cuticle-treated leaves of Eucalyptus haemastoma. Cuticle-treated leaves exhibited a linear response to increasing D, while control leaves exhibited a curvilinear response. The linear and quadratic terms for the control treatment were both significant (P < 0.05), and the linear slope of the treated regression is significant at P < 0.0001.

There was also a qualitative difference in the response of gs, as a function of transpiration, under increasing D, for control and cuticle-treated E. haemastoma leaves (Fig. 3). For control leaves, the three-phase response was observed. Thus, as D increased from low to moderate values (1.4–2.2 kPa), gs remained relatively constant as transpiration increased from 3.6 to 6.4 mmol m−2 s−1 (regime C). As D increased in the mid-range (2.2–2.8 kPa), transpiration remained relatively constant (between 6.4 and 6.52 mmol m−2 s−1) as gs decreased from 297 to 242 mmol m−2 s−1 (regime A). As D exceeded 2.8 kPa, gs and transpiration declined to 211 and 6.1 mmol m−2 s−1, respectively (regime B).

Figure 3.

Stomatal conductance of leaves of Eucalyptus haemastoma shows a three-phase response to E in control leaves, but only a two-phase response in leaves with an increased cuticular conductance.

For solvent-treated leaves, as D increased over the range 1.6–2.9 kPa, all data points occurred within regimes A and B (Fig. 3). Maximum transpiration rate was more than halved in the treated leaves compared to controls, from 6.5 to 3.1 mmol m−2 s−1 (control and treated, respectively). Minimum gs for treated leaves was more than halved compared to controls (from 211 mmol m−2 s−1 in control leaves to 86 mmol m−2 s−1 in treated leaves). Maximum gs for treated leaves was also lower compared to the control leaves (191 and 297 mmol m−2 s−1 for treated and control leaves, respectively). In solvent-treated leaves, as D increased from 1.6 to 2.1 kPa, transpiration remained constant between 3.0 and 3.1 mmol m−2 s−1 as gs declined from 191 to 149 mmol m−2 s−1. Above 2.1 kPa, both gs and transpiration declined with increasing D to 86 and 2.5 mmol m−2 s−1, respectively (Fig. 3).

Over the temperature range 19–40 °C, transpiration (and hence, stomatal conductance as D was constant) showed an approximately parabolic response (Fig. 4a) whereby transpiration was maximal at a temperature of approximately 26–32 °C, and declined at lower or higher temperatures. Over the full temperature range of the experiment, cuticular transpiration rate increased approximately exponentially (Fig. 4a). The ratio of cuticular transpiration to total transpiration increased gradually over the range 18–30 °C but above ca. 32 °C, the rate of increase in the ratio increased (Fig. 4b). This agreed well with the model prediction (Fig. 4c). The concentration of CO2 within the leaf (Ci) remained constant across the entire temperature range used in the second experiment (Fig. 4d).

Figure 4.

(a) Stomatal conductance of leaves of Eucalyptus haemastoma increases to a maximum and then declines as temperature increases at constant D. However, (b) the ratio of Ecut to Etotal increases with increasing temperature at constant D. Similarly, the Eamus and Shanahan model shows an increase in this ratio with increasing temperature at constant D (c), while observations show Ci to remain relatively constant across the full range of temperatures used experimentally (d).

Using the Eamus and Shanahan model with the temperature dependence of gc incorporated, it was possible to model the response of gs to transpiration rate for control leaves (no treatment to the cuticle) and for leaves with gc increased by a factor of 3 (Fig. 5). For the default value of gc, the three-phase response of gs to E was observed, but when gc was increased by a factor of 3, the response curve collapsed to a two-phase curve. This is similar to the experimental results in Fig. 3.

Figure 5.

The Eamus & Shanahan (2002) model predicts changes in stomatal conductance as a function of transpiration rate, for the default and increased gc values. In the default value, the three-phase response is observed, but when gc is increased, stomatal conductance shows a two-phase response.

When gc was decreased by the application of a light, microscope immersion oil to a small patch of leaf of three species, a significant increase (averaging 20%) in stomatal aperture was recorded in five of the six comparisons (3 species × 2 light levels) (Table 1). In contrast, as the VPD of the air above the leaf surface of a R. discolor plant was increased, guard cell turgor declined significantly (Fig. 6).

Table 1.  Change in stomatal apertures following application of microscopic immersion oil to the surface of leaves of three broadleaved annuals
SpeciesLight levelStomatal aperture ± SE (µm) before oil addedStomatal aperture ± SE (µm) after oil added% Change in apertureP value
Commelina communisHigh13.75 ± 0.3113.29 ± 0.42−3.300.380
Moderate11.48 ± 0.3113.68 ± 0.3819.160.001
Commelina cyaneaHigh9.72 ± 0.3711.40 ± 0.3617.280.002
Moderate8.84 ± 0.3411.36 ± 0.3228.51<0.000
Vicia fabaHigh11.76 ± 0.4814.12 ± 0.3820.06<0.000
Moderate11.31 ± 0.4113.12 ± 0.5816.010.001
Figure 6.

Guard cell turgor declines with increasing vapour pressure deficit (VPD) of air above leaves of Rhoeo discolor. Error bars represent standard deviations of turgor.


Stomatal responses to D and the influence of gc on stomatal responses to D

As D increased from 2 to 3.5 kPa, gs declined, as is generally observed (Lange et al. 1971; Schulze 1993; Bunce 1996; Franks et al. 1997). The non-significant increase in gs observed between 1.25 and 2.0 kPa observed here has also been observed on previous occasions (Eamus & Cole 1997; Prior, Eamus & Duff 1997).

The stomata of E. haemastoma showed the three-phase response of stomatal conductance plotted as a function of E, when D was increased to manipulate E (Fig. 3). Such a three-phase response has been observed previously (Monteith 1995; Thomas & Eamus 1999) and has been successfully modelled (Eamus & Shanahan 2002). This mechanistic model is based purely upon feedback behaviour between guard cell turgor and the loss of water from guard cells and epidermal cells, and shows that feed-forward behaviour need not be invoked to explain reduced transpiration rates with increased D. Monteith (1995) showed that such feed-forward behaviour was rare, while Franks et al. (1997) conclude that apparent feed-forward behaviour could be an artefact of experimental conditions.

Sensitivity analysis of the model of Eamus & Shanahan (2002) showed that changes in leaf thickness, the fraction of leaf volume that is air space and the fraction of mesophyll cell wall in contact with air, had minimal impact on model outcomes. However, they predicted that changes in cuticular conductance (gc) would have significant impacts on model behaviour. Thus, when gc was increased, and hence, epidermal transpiration increased, the three-phase response of gc collapsed to two phases and the gs versus E curve shifted to the left. By using a hexane–ethanol solvent mix (Kerstiens 1997), we were able to significantly increase gc (by approximately 250%) and showed that the gs versus E curve shifted to the left and collapsed to a two-phase response, as predicted in the model.

In addition to increasing gc using the solvent treatment on leaves, we also decreased gc of a small region of leaf by covering the surface of a patch of leaf with light, microscope immersion oil. The volume of oil applied and its location in a direction directed away from guard cells was such that its impact was restricted to reducing cuticular transpiration only; it did not reduce stomatal transpiration, which must have increased because stomatal aperture increased, typically by 20%, following the application of the oil (Table 1). From the results of manipulating gc (experiments 1, 3), we conclude that changes in stomatal conductance in response to changes in D are mediated through changes in cuticular transpiration rate, and suggest that transpiration across both the cuticle and through stomata is important in causing the three-phase response of stomata to E. This mechanism is now discussed.

On cuticular conductance and transpiration

Stomatal aperture (and hence, conductance) increases as guard cell turgor increases (Meidner & Bannister 1979; Franks et al. 1995). The cause of the increase in guard cell turgor is water influx to the guard cell, and this is the result of solute accumulation in the guard cell (Raschke 1979). Conversely, a decline in aperture is the result of a decline in turgor because of a decline in the volume of the guard cells which occurs after solute efflux.

Guard cells lose water continually across their external wall and cuticle and also into the substomatal cavity (Franks 2004). To maintain a constant aperture, and hence guard cell turgor, this water is replaced by water influx from an adjacent subsidiary cell, which receives water from adjacent epidermal cells. Water movement between adjacent cells occurs down a water potential gradient.

The Eamus & Shanahan (2002) model makes four predictions about stomatal responses to D. First, increasing D causes a reduction in the ability of the epidermis to supply water to guard cells because of an increase in cuticular transpiration. Consequently, increasing D causes a decline in epidermal cell turgor, and this response should be moderately fast (1 or 2 min because the half-time for equilibration of water across cell membranes is measured in tens of seconds; Eamus & Wilson 1984). Second, the decline in supply of water to the guard cells causes a decline in turgor pressure of the guard cells. Third, increasing cuticular conductance should exacerbate the effect of increasing D such that the stomatal response to increasing D tends towards that observed during soil drought when soil moisture content is limiting and the supply of water to the entire leaf is restricted. Finally, as temperature increases at constant D, gs declines and an increase in the ratio of cuticular to total transpiration rate should occur at constant D because of the strong temperature dependence of gc (Schonherr et al. 1979). Are any of these predictions supported experimentally?

There is strong supporting evidence from earlier studies of the response of epidermal cell turgor to changes in D (prediction 1). For example, Fricke (1997) has shown that a sudden increase in D around a leaf causes a decline in epidermal cell turgor within 2 min, and this is matched by an equally fast decline in gs. Similarly, Mott & Franks (2001) showed an equally rapid decline in epidermal cell turgor in response to increased D. Clearly, the decline in epidermal cell turgor with increased D (Schulze 1993; Fricke 1997; Mott & Franks 2001) supports the first of the three model predictions. More direct evidence in support of the second prediction is the observed decline in guard cell turgor as the VPD of the air above the stomata was increased (Fig. 6). This also agrees with the conclusion of Nonami et al. (1990) who concluded that the decrease in transpiration rate observed in response to increasing D is the result of shortage of water supply to the guard cells from subsidiary cells.

Evidence supporting the third prediction is contained in Fig. 3, where it is shown that the three-phase response to gs was reduced to one or two phases when gc was increased. When the model is run with an increase in gc of 200% (data not shown) or 300% (Fig. 5), a collapse of the three phases to two phases is observed, and thus the impact of the observed increase in gc of approximately 250% resulting from treatment of leaves with the solvent mixture is in full agreement with model predictions. This collapse from three to two and one phases has been observed previously in droughted plants (Thomas & Eamus 1999). Data supporting the fourth prediction is presented in Fig. 4 and in the results of others (Aphalo & Jarvis 1991; Cole 1994; Eamus, Duff & Berryman 1995), where despite a constant D, gs declines with increasing temperature. This increase in temperatures causes increased cuticular transpiration (Fig. 4) and a reduction in water supply to the guard cell. The reduction in gs with increased temperature is not simply a stomatal response to increased Ci because Ci remained constant across the full range of temperatures (Fig. 4).

In conclusion, experimental data derived from several independent methods and a mechanistic model of stomatal responses to D that involves cuticular transpiration and local water potential gradients within the epidermis show close agreement. Consequently, we conclude that a simple feedback loop from increased D to reduced stomatal aperture, via the influence of D on cuticular transpiration, and hence the ability of the epidermis to supply water to guard cells, is sufficient to explain the three-phase response of stomata to increasing D.