Transpiration and stomatal behaviour of Quercus ilex plants during the summer in a Mediterranean carbon dioxide spring

Authors

  • R. Tognetti,

    1. Istituto per l’Agrometeorologia e l’Analisi Ambientale applicata all’Agricoltura, Consiglio Nazionale delle Ricerche (IATA – CNR), P.zzale delle Cascine 18, 50144-Firenze, Italy,,
    2. Department of Botany, University of Dublin, Trinity College, Dublin 2, Ireland,,
    Search for more papers by this author
  • A. Longobucco,

    1. Centro Studi per l’Informatica applicata all’Agricoltura (Ce.S.I.A.), Accademia dei Georgofili, Logge degli Uffizi Corti, Firenze, Italy
    Search for more papers by this author
  • F. Miglietta,

    1. Istituto per l’Agrometeorologia e l’Analisi Ambientale applicata all’Agricoltura, Consiglio Nazionale delle Ricerche (IATA – CNR), P.zzale delle Cascine 18, 50144-Firenze, Italy,,
    Search for more papers by this author
  • A. Raschi

    1. Istituto per l’Agrometeorologia e l’Analisi Ambientale applicata all’Agricoltura, Consiglio Nazionale delle Ricerche (IATA – CNR), P.zzale delle Cascine 18, 50144-Firenze, Italy,,
    Search for more papers by this author

R. Tognetti Fax + 39–55–308910; e-mail: tognetti@sunserver.iata.fi.cnr.it

Abstract

Variations in the water relations and stomatal response of Quercus ilex were analysed under field conditions by comparing trees at two locations in a Mediterranean environment during two consecutive summers (1993 and 1994). We used the heat-pulse velocity technique to estimate transpirational water use of trees during a 5 month period from June to November 1994. At the end of sap flow measurements, the trees were harvested, and the foliage and sapwood area measured. A distinct environmental gradient exists between the two sites with higher atmospheric CO2 concentrations in the proximity of a natural CO2 spring. Trees at the spring site have been growing for generations in elevated atmospheric CO2 concentrations. At both sites, maximum leaf conductance was related to predawn shoot water potential. The effects of water deficits on water relations and whole-plant transpiration during the summer drought were severe. Leaf conductance and water potential recovered after major rainfall in September to predrought values. Sap flow, leaf conductance and predawn water potential decreased in parallel with increases in hydraulic resistance, reaching a minimum in mid-summer. These relationships are in agreement with the hypothesis of the stomatal control of transpiration to prevent desiccation damage but also to avoid ‘runaway embolism’. Trees at the CO2 spring underwent less reduction in hydraulic resistance for a given value of predawn water potential. The decrease in leaf conductance caused by elevated CO2 was limited and tended to be less at high than at low atmospheric vapour pressure deficit. Mean (and diurnal) sap flux were consistently higher in the control site trees than in the CO2 spring trees. The degree of reduction in water use between the two sites varied among the summer periods. The control site trees had consistently higher sap flow at corresponding values of either sapwood cross-sectional area or foliage area. Larger trees displayed smaller differences than smaller trees, between the control and the CO2 spring trees. A strong association between foliage area and sapwood cross-sectional area was found in both the control and the CO2 spring trees, the latter supporting a smaller foliage area at the corresponding sapwood stem cross-sectional area. The specific leaf area (SLA) of the foliage was not influenced by site. The results are discussed in terms of the effects of elevated CO2 on plant water use at the organ and whole-tree scale.

INTRODUCTION

The concentration of carbon dioxide (CO2) in the atmosphere has been increasing from the preindustrial level of ≈ 280μmol mol–1 to the present value of 360 μmol mol–1, and will reach more than twice the preindustrial concentration during the next century (Watson et al. 1990). There is agreement among different climate models that the accumulation of CO2 and other greenhouse gases in the atmosphere could cause an increase in the mean global temperature at the surface of the Earth. Because of the likely increase in transpiration linked to this increase, plant communities of the Mediterranean area may have to face more severe drought conditions, as present rainfall barely meets potential evapotranspiration. Moreover, a concurrent decrease in precipitation at Mediterranean latitudes is also expected (Wigley, Briffa & Jones 1984). The direct effects of elevated CO2 on plant water relations may, therefore, be extremely relevant for Mediterranean forest tree species. In this respect, the response of native Mediterranean plant communities to global change is still poorly documented (Scarascia Mugnozza et al. 1996). Quercus ilex L. (holm oak, evergreen) is the dominant tree species of many mature communities over large areas of the Mediterranean Basin (Romane & Terradas 1992; Damesin & Rambal 1995).

In seedlings, high concentrations of atmospheric CO2 have often been found to improve water status by inducing stomatal closure [see Tyree & Alexander (1993)], but this has not always been observed (Beadle, Jarvis & Neilson 1979; Higginbotham et al. 1985; Nijs, Impens & Behaeghe 1988). In general, a decrease in transpiration and, because of increased rates of photosynthesis, an increase in water use efficiency (Jarvis 1989; Eamus 1991) are expected to take place, although the predictive value of most of the published data is scarce, as most of the elevated CO2 experiments on forest tree species have been relatively short term and carried out with young potted plants, often growing on artificial substrates. An important aspect of drought survival in plants growing in an elevated CO2 atmosphere is the maintenance of a high plant water potential (Rogers et al. 1984; Tolley & Strain 1984). It has been shown that osmotic adjustment (lower solute potential) in leaves of plants exposed to elevated CO2 allows them to maintain higher turgor pressure (Morse et al. 1993; Tognetti et al. 1996a) that may help in sustaining growth and metabolism during drought. On the other hand, a substantial reduction in transpiration may derive from a reduced biomass allocation to branches, and a consequent reduction in total branch leaf area. This kind of response has been observed in some short-term experiments on Eucalyptus seedlings (Duff, Berryman & Eamus 1994) and Populus clones (Ceulemans, Perez-Leroux & Shao 1994), while other experiments showed the opposite response (Rochefort & Bazzaz 1992).

It has recently been demonstrated (Miglietta & Raschi 1993; Körner & Miglietta 1994) that naturally CO2-enriched sites (termed CO2 springs) provide the opportunity to study adult trees exposed throughout their development to an enriched CO2 atmosphere. Several trees of different Mediterranean species growing close by the Bossoleto CO2 spring near Rapolano Terme (central Italy) (Körner & Miglietta 1994; Chaves et al. 1995; Jones et al. 1995; van Gardingen et al. 1995; Tognetti et al. 1996a; Hättenschwiler et al. 1997a) provide an opportunity to investigate long-term responses to concurrent CO2 increase and water stress (as well as to compare different species in their drought tolerance). This natural experiment at CO2 springs is unique and contributes to our understanding of plant responses to elevated CO2, despite the lack of an ‘exact’ control, as the trees have been exposed to elevated CO2 (as well a large range of natural disturbances) for generations.

The investigation described here was part of a project studying the effects of elevated CO2 on water relations and hydraulic architecture of Mediterranean oak species. The ultimate goal is to predict the response of holm oak with respect to a variety of scenarios related to long-term climate change. The aim of the present study was to assess the effects of elevated CO2 on the variability of stomatal response, the water relations and the tree water use in mature Q. ilex trees in natural field conditions in a Mediterranean environment over two consecutive summers. Our hypothesis was that long-term differences in atmospheric CO2 concentration between the high CO2 concentration site and a neighbouring control site would result in improved water use.

MATERIALS AND METHODS

Site description

The natural CO2 spring of Bossoleto (43°17’ N, 11°35’ E and 350 m above sea level) is located near Rapolano Terme (Siena, central Italy); detailed information on the site is given elsewhere (Miglietta et al. 1993; Körner & Miglietta 1994; van Gardingen et al. 1995). The CO2 vents occur both at the bottom and on the flanks of a circular doline; concentration gradients are enhanced under stable (windless) atmospheric conditions. The CO2 concentrations around the crowns of the trees on which the experiment was performed ranged during the day from 500 to 1000 μmol mol–1 with rapid fluctuations. The H2S and SO2 concentrations in the spring are very low and cannot be considered harmful to plants (H. Rennenberg, University of Freiburg, personal communication). A large part of the doline is coppiced forest in which Q. ilex is the main species. A site, 3 km from Bossoleto, with similar morphology, light exposure, soil nutrient availability and depth, plant age and association, was chosen as the control site. The same site has already been used by others as the control site for research conducted in Bossoleto (Chaves et al. 1995; van Gardingen et al. 1997). Scaffolding and/or natural rocks allowed access to leaves on branches (2–5 m above ground level) exposed to direct sunlight for similar periods of the day.

Measurements of leaf conductance and water potential were made on sunny and clear days at 3–5 week intervals from mid-May to the end of September over two consecutive summers, 1993 and 1994, on several trees of Q. ilex, ≈ 11 cm in diameter, and 6 m in height. In Rapolano, the climate is typically Mediterranean with rainfall occurring mainly during autumn and winter. The total rainfall during the study period was 209 mm (total annual rainfall 650 mm) for 1993, and 367 mm (total annual rainfall 791 mm), for 1994. The mean monthly minimum and maximum temperatures ranged from – 1·1/2·4 °C (February 1993/1994) and 8·8/9·8 °C (January 1993/1994) to 18·4/19·4 °C (August 1993/July 1994) and 31·2/32·3 °C (August 1993/1994). Both summers were very dry and hot. The two sites did not display relevant differences in rainfall.

Stomatal response and tree water relations

Diurnal courses of leaf conductance (gl) and transpiration (E) were followed on the abaxial surface at 2 h intervals from predawn to sunset, using a null balance steady-state porometer (LI-1600, Li-Cor Inc., Lincoln, Nebraska, USA). Measurements were made on four different trees per site and per species. Six fully developed leaves (last flush) were measured on each tree on each sampling occasion. The natural inclination and azimuth of the leaves were maintained during the measurements. For each measurement, the temperature inside the cuvette, leaf temperature, relative humidity and incoming photosynthetic photon flux density (PPFD) were recorded. Precautions were taken to avoid large differences between external environmental conditions and those inside the cuvette. Cuvette overheating was minimized by shading the cuvette between measurements. The porometer temperature and humidity sensors were calibrated and checked regularly. Leaf conductances were expressed in molar units to account for differences in temperature and atmospheric pressure resulting from altitude. Changes in leaf conductance were considered to reflect changes in stomatal conductance, on the assumption that boundary layer conductance inside the cuvette was constant and large.

Daily courses of xylem water potential (Ψ) were determined using a pressure chamber (PMS-100, P.M.S. Instruments Co., Corvallis, Oregon, USA) at 2–3 h intervals, from predawn to sunset, in parallel with the diurnal determinations of transpiration rate and leaf conductance. Three to six terminal shoots per tree were chosen close to the leaves on which leaf conductance had been measured. The absolute difference between the minimum water potential (Ψm), generally recorded at midday, and the predawn water potential (Ψpd) was calculated and termed ΔΨ.

Transpiration at predawn was assumed to be zero because dew condensation on leaves before sunrise was very common, even during mid-summer. Hydraulic resistance on each tree sampled was calculated as the slope of the relationship between the xylem water potential and the transpiration rate per unit leaf area (Hinckley & Bruckerhoff 1975; Brough, Jones & Grace 1986; Sala & Tenhunen 1994; Tognetti et al. 1996b). Diurnal hysteresis was small. Only regressions significant at P < 0·01 were used for hydraulic resistance estimates.

For comparative purposes, curves describing changes in measured maximum leaf conductance (approximately equal to stomatal conductance) with respect to atmospheric vapour pressure deficit were obtained from the entire data set (Lange et al. 1987). The maximum daily leaf conductance at light saturation (gl max) was defined as the average of the highest values of leaf conductance for each day. Data from diurnal courses indicated that the time when the maximum daily light-saturated leaf conductances generally occurred was between 0700 and 0900 h solar time. Regressions of stomatal response to atmospheric vapour pressure deficit were constructed by pooling values obtained from trees sampled at each site (each data pair corresponding to individual leaves).

Sap flow measurements

Over the entire period from June to mid-November 1994, measurements of half-hourly sap velocities were made simultaneously on a total of 12 Q. ilex trees (six at each site) by the heat-pulse technique (Raschi et al. 1995; Tognetti et al. 1996a,b). On each tree, two sets of probes were implanted opposite one another and monitored with a data logger. Sap velocity sensor units (Horticulture and Food Research Institute, Palmerston North, New Zealand) were deployed at breast height on each tree; heating probes penetrated the xylem to a depth of 30–35 mm, whereas two thermistor probes were inserted into the xylem 20–25 mm beneath the cambium. Thermistor probes were sited 10 mm above and 5 mm below each heating probe. Heating and thermistor probes were vertically aligned. At 30 min intervals, 5 s heat pulses were automatically triggered by the logger. An increment core (5 mm diameter) was taken near to where the probes were implanted to determine, by colour and texture, the approximate inner and outer boundaries of the sapwood and its volume fractions of wood and water (more accurate measurements of sapwood area were made on wood discs after felling, see below). The calculation of sap velocity from heat-pulse velocity was made after Swanson (1962) and included correction for the effect of wounding (Swanson & Whitfield 1981). Sap velocity calculations were restricted to daytime (sunrise to sunset). Marshall’s (1958) equation was used to convert heat-pulse velocity to sap flow.

Destructive measurements

After completion of the transpiration measurements all sample trees were felled. The total leaf mass from each tree was harvested and weighed, after reaching constant mass in the oven at 70 °C. Before drying, a subsample of leaves stripped from each tree was immediately fed through a planimeter (LI-3000, Li-Cor Inc.) and used to obtain the total foliage area from the total mass by regression analysis. The specific leaf area (SLA) was obtained as the ratio between the total foliage area and the total foliage dry mass. The SLA was also calculated on leaf disks cut from leaves randomly chosen (30 leaves per plant, three discs per leaf). Stem discs were cut at probe implantation height for measurements of the sapwood cross-sectional area. Measurements were made of sapwood thickness and heartwood diameter at up to eight points around each wood disc were determined.

RESULTS

Seasonal patterns of xylem water potential and leaf conductance

As both years were characterized by severe summer drought, the trees showed large reductions in leaf conductance and water potential during the summer, with particularly low values in 1993 (lower rainfall) (Fig. 1). The predawn shoot water potential was correlated with rainfall. In both years, Ψpd remained close to zero in May and June. At the beginning of July, a sharp decrease in Ψpd was observed at both sites. After the first major rainfall events, at the beginning of September, Ψpd recovered to ≈–0·5MPa. Overall, significant differences in Ψpd between the two sites were not observed. A consequence of the seasonal patterns in Ψpd and Ψm (data not shown) was the observation that ΔΨ showed two maxima: before and after drought (Fig. 1). In late September, ΔΨ increased to values similar to those found in May and June. During drought development, ΔΨ decreased strongly with decreasing Ψpd.

Figure 1.

. Seasonal course of maximum leaf conductance (gl max), predawn shoot water potential (Ψpd) and differences between mean predawn and minimum shoot water potentials (ΔΨ) measured in the sun crown of Quercus ilex trees at the CO2 spring site (closed symbols and continuous line) and at the control site (open symbols and dashed line) during clear days in summer 1993 and 1994. Vertical error bars (visible when larger than symbols) indicate two standard errors of the mean. Weekly precipitation measured during the study period is indicated by vertical bars.

The maximum leaf conductance in sun leaves varied in parallel with Ψpd (Fig. 1). During the wet period (May and September), there was a small difference between the CO2 spring and the control site, gl max being higher in the latter. At both sites, at the end of the drought period, gl max recovered to predrought values.

The relationships between ΔΨ and Ψpd and between gl max and Ψpd (1993 and 1994 pooled data) are described in Fig. 2.

Figure 2.

. Changes in maximum leaf conductance (gl max) and ΔΨ with changes in predawn shoot water potential (Ψpd) of Quercus ilex at the CO2 spring site (closed symbols) and at the control site (open symbols) during the study period (1993 and 1994 pooled data). Vertical and horizontal bars (visible when larger than symbols) indicate two standard errors of the mean. Each line in the upper panel represents a second order polynomial equation fitted to the data (P < 0·001) for trees at the CO2 spring site (R2 = 0·76) and the control site (R2 = 0·83), respectively. Each line in the lower panel represents a second order polynomial equation fitted to the data (P < 0·001) for trees at the CO2 spring site (R2 = 0·75) and the control site (R2 = 0·87), respectively.

Diurnal patterns in leaf conductance and its changes with atmospheric vapour pressure deficit

Leaf conductance was always relatively high in the morning and decreased gradually during the day (Fig. 3), at both sites (a representative selection of daily trends for the 2 years is reported). Midday depression was evident at the beginning of the summer. By mid-August, gl was greatly reduced and stomata were almost completely closed, particularly during the afternoon. After the rain in September, gl of trees at both sites increased. Leaf conductance was higher in the control trees than in the CO2 spring trees during the period of mild water stress.

Figure 3.

. Diurnal courses of leaf conductance (gl) of leaves from the sun crown of Quercus ilex trees at the CO2 spring site (closed symbols and continuous line) and at the control site (open symbols and dashed line) during two consecutive summers, 1993 and 1994. A representative selection of daily trends for the 2 years is reported. Vertical error bars (visible when larger than symbols) indicate two standard errors of the mean.

Over the long term (months), gl max decreased with increasing vapour pressure deficit in trees both when grown and measured at ambient CO2 and when grown and measured at elevated CO2 (Fig. 4). The response was relatively steeper for the control trees than for the trees at the CO2 spring site (1993 and 1994 pooled data).

Figure 4.

. Changes in maximum leaf conductance (gl max) with changes in atmospheric vapour pressure deficit (VPD) of Quercus ilex trees at the CO2 spring site (closed symbols and continuous line) and at the control site (open symbols and dashed line) during the study period (1993 and 1994 pooled data). Vertical and horizontal error bars (visible when larger than symbols) indicate two standard errors of the mean. Each line represents a second order polynomial equation fitted to the data (P < 0·001) for trees at the CO2 spring site (R2 = 0·86) and the control site (R2 = 0·98), respectively.

Seasonal changes in hydraulic resistance

A large increase in hydraulic resistance (Fig. 5) occurred at both sites in mid-summer (particularly in 1993), coinciding with large reductions in leaf conductance. In 1993, hydraulic resistance values reached ≈1·5Mpa m2 s mmol–1 in August, whereas, in 1994, values were ≈1 Mpa m2 s mmol–1, and in 1994 were consistently higher at the control site than at the CO2 spring site. At the end of September, hydraulic resistance recovered to predrought values.

Figure 5.

. Seasonal variation of hydraulic resistance estimated from cuvette transpiration measurements and average shoot water potential of Quercus ilex trees at the CO2 spring site (closed symbols and continuous line) and at the control site (open symbols and dashed line) during two consecutive summers, 1993 and 1994. Vertical error bars (visible when larger than symbols) indicate two standard errors of the mean.

The relationships between gl max and hydraulic resistance and between Ψpd and hydraulic resistance (1993 and 1994 pooled data) are shown in Fig. 6. At corresponding low Ψpd, CO2 spring trees had relatively smaller values of hydraulic resistance than the control site trees.

Figure 6.

. Changes in maximum leaf conductance (gl max) and predawn shoot water potential (Ψpd) with changes in hydraulic resistance of Quercus ilex trees at the CO2 spring site (closed symbols and continuous line) and at the control site (open symbols and dashed line) during the study period (1993 and 1994 pooled data). Vertical and horizontal error bars (visible when larger than symbols) indicate two standard errors of the mean. Each line in the upper panel represents a second order polynomial equation fitted to the data (P < 0·001) for trees at the CO2 spring site (R2 = 0·77) and the control site (R2 = 0·90), respectively. Each line in the lower panel represents a second order polynomial equation fitted to the data (P < 0·001) for trees at the CO2 spring site (R2 = 0·93) and the control site (R2 = 0·94), respectively.

Sap flow, foliage and sapwood area

Various tree characteristics relating to the interpretation of sap flow are summarized in Table 1. The mean daily water use decreased markedly throughout the rainless period in larger trees and less evidently in smaller trees, regardless of site (Fig. 7). The peak daily water use was recorded before (June) the onset of the mid-summer drought (August) and after the rainfall events of late summer (September) which signalled recovery from soil water deficit, suggesting that soil water recharge resulted in a rapid increase in transpiration rate. Daily water use started decreasing by the end of October as temperatures decreased and the photoperiod shortened. Unstable transpiration rates throughout the experiment (but consistently less during the dry period) are presumably the result of corresponding variability in incident radiation and vapour pressure deficit. There was a considerable variation in the mean daily sap flow among the trees sampled related to tree dimensions, while the mean daily sap velocity did not show a systematic relationship to stem diameter (Table 1). The mean sap flux and velocity were consistently higher in the control trees than in the CO2 spring trees.

Table 1.  . Parameter values for all Quercus ilex trees sampled for sap flow at the CO2 spring site and the control site Thumbnail image of
Figure 7.

. Daily daytime sap flow over the measurement period (1994) for all the sampled Quercus ilex trees at the control site (upper panel, dashed lines) and the CO2 spring site (lower panel, continuous lines). Interruptions in the continuity of the lines result from failure, and subsequent replacement, of probes.

Diurnal sap flow (Fig. 8) of CO2 spring trees was consistently lower than that of control trees, particularly during the relatively wet period. Sap flow exhibited the typical parabolic pattern, peaking in mid-morning before beginning a steady decline throughout the remainder of the day. The effect of water stress was marked and partially masked that of CO2 and, as a result, at peak water stress the difference between CO2 spring and control trees was smaller.

Figure 8.

. Diurnal trends of sap flow representative of prestress (upper panel) and peak stress (lower panel) period (DOY = day of year). Data are half-hourly averages from sunrise to sunset of sap flow in all measured Quercus ilex trees at the control site (open symbols) and the CO2 spring site (closed symbols). Vertical and horizontal error bars (visible when larger than symbols) indicate two standard errors of the mean.

After ln transformation to linearize the data (see also the other relationships), the mean daily sap flow was strongly related to both sapwood area (at probe implantation height) and foliage area (Fig. 9). The control trees had consistently higher sap flow at corresponding values of either sapwood or foliage area than the CO2 spring trees, but larger trees showed smaller differences.

Figure 9.

. Relationship between mean daily daytime sap flow and sapwood area and foliage area, and between foliage area and sapwood area for Quercus ilex trees at the control site (open symbols) and the CO2 spring site (closed symbols). Lines (dashed for control trees and continuous for CO2 spring trees) were fitted by least squares regression for predictive purposes, following ln transformation of both variables to linearize the data.

There was a strong association between foliage area and sapwood area in both control and CO2 spring trees (Fig. 9), the latter supporting a smaller foliage area at the corresponding sapwood area. Sapwood areas were reliably correlated to stem cross-sectional areas (over bark), but in the opposite direction, i.e. the CO2 spring trees had a larger sapwood area at the corresponding stem cross-sectional area than the control trees (data not shown).

The average SLA was not influenced by site, and was similar in the control and the CO2 spring trees (5·97±0·69 m2 kg–1 and 6·11 ± 0·35 m2 kg–1, respectively).

DISCUSSION

Seasonal and diurnal water relations during summer drought

Seasonal and diurnal patterns of leaf conductance and xylem water potential at the two sites were similar to those observed previously in the same and other western Mediterranean sclerophyllous species (Tenhunen et al. 1987; Rhizopoulou & Mitrakos 1990; Romane & Terradas 1992; Sala & Tenhunen 1994; Castell, Terradas & Tenhunen 1994; Damesin & Rambal 1995).

As in other Mediterranean species, stomatal closure in Q. ilex constitutes a mechanism to cope with diurnal and seasonal water deficits (Aussenac & Valette 1982; Lange, Tenhunen & Braun 1982; Archerar & Rambal 1992; Castell et al. 1994; Sala & Tenhunen 1994; Damesin & Rambal 1995).

The progressive stomatal closure with decreasing predawn water potential indicates conservative water use and also explains the decrease in ΔΨ during drought (Damesin & Rambal 1995). Yet, the minimum predawn water potential on trees at Rapolano (particularly in 1993) was between – 4 and – 6 MPa, which is similar to that measured on Mediterranean Quercus species in very dry years or on sites close to desert areas (Rambal & Debussche 1995). Extreme stress conditions were, in fact, evidenced by yellowing leaves during August (1993), probably indicating irreversible cell damage (Kyriakopoulos & Larcher 1976). Drought-induced senescence and shedding of leaves is a known means to avoid marked drought stress (Kramer 1980). Overall, the plants recovered to high water potentials and leaf conductances after the return of the September rains, allowing continuation of annual net production (Damesin & Rambal 1995).

At the same sites in late June 1993, Tognetti et al. (1996a) found an osmotic potential of ≈– 3·35 MPa at turgor loss point in Q. ilex trees growing at the CO2 spring and ≈ 3·05 MPa in the control trees. In this study, predawn water potentials were near or below this critical value when the drought was at its peak (July and August). Although water loss was still measurable, stomata opened only for a short period in the early morning. However, strong reductions in stomatal conductance occurred when water potential approached the turgor loss point (Hinckley et al. 1980; Rhizopoulou & Mitrakos 1990; Sala & Tenhunen 1994), coinciding with large decreases in predawn water potential. Osmotic adjustment can be an important complementary mechanism of drought resistance that operates to enable the stomata to open in conditions of severe soil water stress, lowering the water potential at which closure occurs (Rhizopoulou & Mitrakos 1990; Terradas & Savé 1992).

The heat-pulse velocity technique has already been applied successfully on these and other Quercus species (Miller, Vavrina & Christensen 1980; Borghetti et al. 1993; Raschi et al. 1995; Tognetti et al. 1996a,b). Sap velocity and sap flow decreased in parallel with increases in hydraulic resistance, reaching a minimum in mid-summer. Diffuse xylem embolism (up to 80% in branches) occurred in trees at Rapolano at this time (Tognetti et al. 1996a; Tognetti & Raschi unpublished results). The relationships between decreases in stomatal conductance, predawn water potential and sap flow, and increases in hydraulic resistance are consistent with the suggestion by Tyree (1989) that an important role of stomatal control of transpiration is not only to prevent desiccation damage but also to avoid ‘runaway embolism’, despite a lot of xylem redundancy (Jones & Sutherland 1991; Borghetti et al. 1993). The CO2 spring trees had less reduction in hydraulic resistance for a given value of predawn water potential than the control trees. This might confer the capacity to prolong photosynthesis during dry periods.

Mean peak sap flows declined progressively in concert with increasing drought conditions and reached a minimum when water stress was at its maximum (mid-summer). With the onset of late summer rainfall, mean peak sap flows recovered to prestress values, regardless of site. This seasonal pattern, of progressive decline and recovery, may be attributed to increasing and decreasing soil water deficit. The presence of shallow bedrock at both sites may also have prevented trees abstracting water from deeper soil water reserves, so that the trees responded rapidly to soil water deficit and recharge of the upper horizons, which have a low water-holding capacity. On the other hand, the distribution of roots in this kind of substrate is difficult to assess and, although, unlikely, differential access to soil water cannot be excluded.

Effects of high CO2 concentration

It is a general, but not a universal, observation that leaf transpiration is reduced in elevated CO2. However, the majority of measurements published so far are from plants grown in controlled-environment chambers, often in pots, and were made at low or moderate leaf-to-air water vapour pressure deficit. Because of the apparent species specificity of response, predictions of the effects of elevated atmospheric CO2 on leaf conductance, particularly of trees, remain uncertain.

In particular, there is relatively little information on how the sensitivity of stomatal conductance to water vapour pressure deficit is affected by either short-term or long-term exposure to elevated CO2. Moreover, the mechanism by which stomata sense humidity is still unclear (Meinzer 1993). Recently, evidence for no stomatal responses to elevated CO2 in tall trees has been reported by several authors (Barton, Lee & Jarvis 1993; Dufrene, Pontailler & Saugier 1993; Ellsworth et al. 1995; Teskey 1995; Körner & Würth 1996; Tognetti et al. 1996a), irrespective of the enrichment method.

Hollinger (1987) found a significantly smaller relative decrease in conductance with increasing leaf-to-air vapour pressure deficit in leaves of two tree species grown and measured in elevated CO2. Leaf conductance in Q. ilex was higher in the control trees than in the CO2 spring trees in the early morning at low vapour pressure deficit, but the difference was small, and this observation was less evident at the peak of water stress in mid-summer. Nevertheless, our data indicate that the decrease in leaf conductance caused by elevated CO2 may be less evident at high atmospheric vapour pressure deficit over the long term. Our observations are consistent with those of Bunce (1993) for soybean (grown outdoors) and orchard grass.

Surprisingly, SLA was not influenced by elevated CO2, in contrast to the reports of many authors [see Ceulemans & Mousseau (1994)]. Again, most studies report data from experiments carried out on seedlings and/or in controlled environments. A different behaviour may be expected by trees in the field, exposed to elevated CO2 for generations (Jones et al. 1995).

Overall, the sap flow measurements showed that water flux from the trees at the CO2 spring was usually less than that from the trees at the control site, but the difference was small. This finding is consistent with the established effects of CO2 on plant water use at the organ and whole-plant scale (Rogers & Dahlman 1993; Bremer, Ham & Owensby 1996; Senock et al. 1996; Dugas, Prior & Rogers 1997). The reduction in the mean daily sap flow was more evident in trees with small foliage and sapwood areas. Half-hourly whole-plant transpiration was larger for trees grown at the control site. However, the degree of reduction in water use between sites varied among the summer periods. There was a small CO2 effect during the peak of water stress in mid-summer, and the degree of reduction caused by drought, that occurred for both the control and the CO2 spring trees because of declining soil water content, was much larger than that attributable to CO2.

Control trees had a consistently larger foliage area at the corresponding sapwood area than CO2 spring trees. In many species, elevated CO2 concentrations have been found to stimulate an increase in transpiring surface area usually associated with the increase in tree size (Ceulemans & Mousseau 1994). However, the majority of experiments have been carried out in conditions of unlimited nutrient and water supply. Chaudhuri, Kirkham & Kanemasu (1990) reported an increased transpiring surface area in wheat grown in elevated CO2, but under free-air CO2 enrichment the same species did not show any increase in leaf area per culm (Senock et al. 1996). The effect of a reduced transpiring surface (lower foliage area) in trees (particularly if of small size) at the CO2 spring site might be equally, if not more, effective than stomatal closure in reducing transpiration and plant water use under elevated CO2.

Hättenschwiler et al. (1997a) found that the regeneration phase of Q. ilex can be accelerated in CO2-enriched atmospheres, while stimulation responses are much less evident when trees are mature; the positive effect of elevated CO2 was relatively larger in years with a dry spring. Hättenschwiler et al. (1997b), studying morphological adjustments to elevated CO2 in mature Q. ilex growing around the CO2 spring of Bossoleto, concluded that the observed enhancement of stem biomass production, but with lower leaf area, may represent an effective mechanism for increased water use efficiency of trees growing in a CO2-enriched atmosphere. These measurements were made in an area of the same CO2 spring, characterized by similar CO2 enrichment but with a different microclimate. Chaves et al. (1995) also reported higher water use efficiency of Q. ilex trees exposed to elevated CO2 for life during the warm hours of the day.

In elevated global atmospheric CO2, the regeneration phase of Q. ilex may be better able to withstand periods of drought than in current ambient CO2 concentrations. This may result in changes in community composition because of varying species response to elevated CO2 and, especially in Mediterranean areas, to enhanced temperature and more severe drought conditions, as forecast by General Circulation Models. A reduction in water usage, which will conserve soil water, might prolong physiological activity during periodic drought. On the other hand, if this reduction in water use is restricted to the smaller trees of a particular group of species, the effect on the water balance of a Mediterranean-type forest will be relatively minor. We may conclude that whole-tree measurements can provide new insights into the consequences of elevated CO2 for transpiration, if coupled with leaf- and ecosystem-scale data.

Acknowledgements

The research was in part supported by the E.U., programme ENVIRONMENT, contract EV5 V-CT93–0093. We thank Jon D. Johnson for helpful discussion.

Ancillary