Introduction

Periods of water deficit occur seasonally in regions with Mediterranean climates. In the past, dryland pasture production in these regions has mainly relied on annual plants that can quickly grow and reproduce during the wet season and survive the dry season as seed, thus avoiding severe drought (Turner 2004). However, perennial plants are needed in many cases to improve the sustainability of agricultural systems and alleviate problems such as dryland salinity (Cocks 2001). Perennials need to be able to resist or tolerate periods of water deficit to survive. Medicago sativa L. (lucerne, alfalfa) is the most productive and widely used perennial forage legume in low rainfall Mediterranean climates (<450 mm mean annual rainfall). The resistance of M. sativa to short growing seasons, and long and intense periods of water deficit is related to its ability to access water from the subsoil and its ability to remain dormant when water supply is limited (Sheaffer et al. 1988). Few other commercial forage options are currently available that are suitable in these climates.

Dorycnium hirsutum (L.) Ser. and Dorycnium rectum (L.) Ser. are two perennial legumes that could provide another forage option in regions with Mediterranean climates (Bennett 2002; Dear et al. 2003). Both species originate from Mediterranean regions of Europe, Africa and East Asia and could be suitably adapted to regions with similar climatic conditions (Bell et al. 2007). D. hirsutum has persisted well under drought conditions in Australia (Lane et al. 2004; Bell et al. 2005) and New Zealand (Wills 1983; Sheppard and Douglas 1986; Wills et al. 1989; Douglas et al. 1996), but the extent of its adaptation to water-limiting conditions is unknown. D. hirsutum is able to extract water from deep soil layers during periods of low rainfall (Bell et al. 2006), but a range of other physiological and morphological adaptations to drought may be important in providing tolerance to water deficit. D. rectum appears to be less successful in drought-prone environments. Survival of D. rectum seedlings was inferior to D. hirsutum under water-limited conditions during its year of establishment (Bell et al. 2005), and anecdotally D. rectum is thought to occur mostly on moist sites or where water is readily available (Demiriz 1970). The rooting depth and capacity of D. rectum to extract water from deep soil layers is unknown. The apparent lower tolerance of D. rectum seedling than D. hirsutum seedlings to drought conditions could be due to differences in root morphology (Bell 2005), but other strategies associated with tolerance of water deficit have not yet been investigated.

While the deep root systems of M. sativa and D. hirsutum are important adaptations to periods of low water supply, a range of other morphological and physiological adaptations to avoid or tolerate water deficit may play a role. These include [adapted from Begg and Turner (1976)]:

  • Regulation of water use by stomatal closure in response to declining leaf turgor. This enables plants to photosynthesise when evaporative demand is low in the morning and evening, and conserve water when evaporative demand is high during the middle of the day.

  • Osmotic adjustment via the net accumulation of solutes in the leaf under water deficit. This lowers leaf osmotic potential and contributes to the maintenance of leaf turgor.

  • Minimisation of leaf area to reduce the rate of water loss from the soil. This slows the development of water deficit and can involve leaf shedding or accelerated senescence of older leaves, leaf movement to reduce incident radiation, and leaf flagging or rolling to reduce effective leaf area.

  • Formation of trichomes or leaf hairs and cuticular waxes. These can increase reflection of radiation and reduce water conductance through the leaf boundary layer, thus reducing water loss and moderating leaf temperature.

This experiment is the first to compare plant responses to increasing water deficit among D. rectum, D. hirsutum and M. sativa. Plants were grown in a set soil volume to reduce the impact of dissimilar root morphologies and investigate plant water relations in isolation from these differences. It was hypothesised that M. sativa and D. hirsutum would possess leaf physiological and morphological adaptations such as stomatal closure, osmotic adjustment and leaf adjustments associated with resistance to plant water deficit, while the expression of these would be inferior in D. rectum.

Materials and methods

Experimental design

The experiment was carried out in the glasshouses at the University of Western Australia, Perth, Australia (31°59′S, 115°53′E). Plants were grown in 7.9 L free-draining pots (1 m deep, 10 cm diameter) which were chosen to be large enough to slow soil drying, but would enable plant water relations to be investigated in isolation of root morphological differences between species. In each pot 400 g of coarse gravel was placed in the bottom and 11 kg of air-dried sandy clay loam soil, collected from CSIRO Yallanbee Research Station, Bakers Hill, Western Australia (31°45′S, 116°29′E), added above the gravel. Soil was allowed to settle to a bulk density of 1.48 g cm−3. The moisture retention characteristics of the soil were determined using the pressure plate technique (Moore et al. 2004). The soil was packed into rings at the same bulk density to that in the pots and the gravimetric water content corresponding to 0, −10, −100 and −1,500 kPa determined (Table 1). Total available water content (i.e. water content between −10 and −1,500 kPa) was 13.8% w/w.

Table 1 Soil water characteristics of sandy clay loam soil used in the glasshouse experiment, as determined by the pressure plate method (n = 4)
Full size table

Basal nutrients were applied per kg of oven-dried soil as follows: 14 mg kg−1 MgCl2; 133 mg kg−1 Ca(H2PO4); 200 mg kg−1 K2SO4; 10 mg kg−1 MnSO4; 10 mg kg−1 ZnSO4; 2 mg kg−1 CuSO4; 67 μg kg−1 H3BO3; 33 μg kg−1 CoSO4 and 17 μg kg−1 Na2MoO4. The nutrients were added in a solution to the soil surface before sowing. Appropriate strains of root nodule bacteria (AL for M. sativa and WSM 1293 for Dorycnium) were irrigated onto the soil surface 3 days after sowing. A small amount of nitrogen (100 mg kg−1 NH4NO3) was also added to aid early growth of seedlings before effective nodules had been formed. Glasshouse air temperatures throughout the experimental period are provided in Fig. 1.

Fig. 1
figure 1

Maximum (filled circle) and minimum (open circle) air temperatures in the glasshouse during the experimental period. Vertical lines indicate the initiation of water-stressed treatments (A—25 November), completion of physiological measurements (B—17 December) and rewatering of pots (C—29 December)

Full size image

Ten pots each were sown with M. sativa L., D. hirsutum (L.) Ser. and D. rectum (L.) Ser. with five replicate pots of each species later either water-stressed or well watered. Pots were allocated to the glasshouse bench in a split-plot design. Ten seeds were sown per pot and seedlings were thinned to one plant per pot approximately 4 weeks after sowing.

An Australian commercially available cultivar (Sceptre) was used for M. sativa, while seed of accession TAS 1002 of D. hirsutum and accession TAS 1274 of D. rectum (E. Hall, Tasmanian Institute of Agricultural Research) were obtained from the Trifolium Genetic Resources Centre, Department of Agriculture and Food, Western Australia. Sowing of seeds was staggered, based on data from Bell (2005), in an attempt to achieve similar plant size at the initiation of the drought treatments; 12 July for D. hirsutum, 19 July for D. rectum and 9 August for M. sativa. Plants were watered to field capacity three times a week until the root system of all plants had fully explored the soil in the pot; spare pots of each species were examined to ensure this was the case. All pots were watered to field capacity on the afternoon of 25 November 2004 and watering then ceased for the water-stressed plants. Well-watered control pots were weighed and watered to weight (field capacity) twice a week, and in the afternoon before physiological measurements were carried out. Watering was withheld from water-stressed plants for a further 2 weeks after physiological measurements were completed (from 21 to 35 days after initiation of water-stress) and pots were rewatered to field capacity and plant survival recorded after a further 10 days.

Physiological measurements

Physiological measurements were carried out on water-stressed and well-watered plants 0, 6, 10, 13, 17 and 21 days after water was withheld. Pots were weighed and the corresponding soil water content (SWC) determined. Leaf water potential was measured predawn (0300–0500 h) and at midday (1200–1400 h) in a pressure chamber (Soilmoisture Equipment Corp., Santa Barbara, CA, USA) on petioles for lucerne and stems of Dorycnium species (petioles were too short). Samples were immediately placed in a water-tight vial, snap frozen on dry ice and stored in a freezer at −20°C. Samples were later thawed, sap expressed using a leaf press and sap osmolality measured using a freezing point osmometer, which was calibrated against 50 and 850 mOsm kg−1 standard solutions (Fiske Associates, Norwood, MA, USA). Osmotic potential (π) of samples was then calculated from osmolality (Eq. 1).

$$ \pi = \frac{{2.447 \times {\hbox{osmolality}}}} {{1,000}} $$
(1)

Osmotic adjustment was calculated as the difference in π at full turgor (π sat; Eq. 2; i.e. 100% relative leaf water content) between water-stressed and well-watered plants according to the method of Ludlow et al. (1983). Relative leaf water content was determined based on the method outlined by Turner (1981). Whole leaves were removed adjacent to those used for water potential measurements and their fresh mass (LeafFW) measured immediately. Turgid mass (LeafTW) was measured after whole leaflets had been floated on deionised water overnight (approximately 16 h) in a petri dish placed in a laboratory with ambient air temperature of 25°C. Dry mass (LeafDW) was measured after samples were dried in a 70°C oven for 2 days.

$$ \pi _{{\hbox{sat}}} = \pi \times \frac{{{\hbox{Leaf}}_{{\hbox{FW}}} - {\hbox{Leaf}}_{{\hbox{DW}}} }} {{{\hbox{Leaf}}_{{\hbox{TW}}} - {\hbox{Mass}}_{{\hbox{DW}}} }} $$
(2)

Measurements of CO2 and H2O exchange were carried out between 0900 and 1100 h using a LI−6400 with red/blue LED light source (LI-COR, Lincoln, NE, USA). Fully expanded leaves were enclosed in the chamber for approximately 2 min and allowed to equilibrate. The light level was set at 1,500 μmol m−2 s−1, flow rate at 500 μmol s−1, [CO2] to 400 μmol mol−1, leaf temperature to 25°C and relative humidity was maintained between 33 and 42%. After measurements were logged, leaves were marked and leaf water potential at midday was taken from or adjacent to this leaf. Leaf area within the cuvette was determined using a WinRHIZO scanner and software (Régent Instruments Inc., Quebec City, QC, Canada) after turgid leaf weights were obtained. Specific leaf area was calculated for these leaves as the leaf area per unit of dry matter (m2 kg−1). Photosynthetic parameters were calculated on a leaf area and dry mass basis and assuming equal distribution of stomata on both leaf surfaces.

Statistical analysis

Water relations measurements, including leaf water potential, osmotic potential and π sat, and photosynthetic water use efficiency (WUE) (A/gS w) and the ratio of internal and external CO2 concentration (C i/C a) were subjected to an analysis of variance using Genstat version 7 (Lawes Agricultural Trust, Rothamsted, UK) with the main effects of species, treatment and time investigated as factors. Species differences in SWC and soil water depletion rate over time were subject to an analysis of variance for droughted plants only. Relationships between photosynthetic rate and stomatal conductance were investigated against leaf water potential (the main determining factor for these processes), but significant regression relationships were not obtained.

Results

Soil water depletion

Depletion of soil water by all species was rapid at SWCs >8% w/w, with more than 70 g of water removed from each pot each day (Fig. 2a). This rate was then reduced at day 13, and at day 21 soil water depletion was the same as for bare soil (data not presented). This low rate (<15 g day−1) continued in all species after day 21 until plants were rewatered to check for survival. D. rectum began with higher SWC than D. hirsutum and this continued at all times during the experimental period (Fig. 2b), even though soil water depletion was similar (Fig. 2a). Water depletion by M. sativa was higher than D. hirsutum and D. rectum until day 10. SWC of well-watered plants was maintained above 60% of field capacity and was >75% of field capacity at the time of all physiological measurements.

Fig. 2
figure 2

Changes in soil water depletion rate (a) and the gravimetric SWC (b) (mean ± SEM, n = 5) in well-watered (solid) and water-stressed (open) pots containing M. sativa (circles), D. hirsutum (squares) and D. rectum (upward triangles) after the initiation of water-stress treatments. Dashed lines indicate SWC at −10 and −1,500 kPa

Full size image

Plant water status

Leaf water potentials were similar for M. sativa and D. hirsutum as SWC declined (Fig. 3a, b). Predawn leaf water potentials remained constant (>−0.7 MPa) until after day 10 (Fig. 3). In M. sativa and D. hirsutum, midday leaf water potential was 1.0 MPa lower than predawn measurements on average, and began to decline by day 10. On day 21, when SWC was <4% w/w (−1,500 kPa), both predawn leaf water potentials of M. sativa and D. hirsutum reached <−3.0 MPa. Midday leaf water potentials for all species on day 21 were below −4.0 MPa (the minimum of the pressure chamber used) or leaves had senesced. D. rectum showed similar leaf water potentials at predawn and midday and plants did not display the same decline in plant water status as observed for D. hirsutum and M. sativa (Fig. 3c). Although SWC was higher at each measurement for D. rectum, it appeared that D. rectum leaf water potentials were maintained until lower levels of soil water (<8% w/w at midday and <6% w/w predawn). For example, D. rectum midday leaf water potentials were −1.6 MPa on day 17 when SWC was 6.5% w/w, while on day 13, D. hirsutum at the same SWC (6.3% w/w) had a midday leaf water potential of −2.4 MPa. A large amount of variation existed between leaf water potentials measured for D. rectum plants as their water status declined, with some plants maintaining water status while others had a rapid drop in leaf water potential.

Fig. 3
figure 3

Predawn (upward triangles) and midday (circles) leaf water potential (a–c) and osmotic potential (d–f) of well-watered (closed) and water-stressed (open) plants of M. sativa, D. hirsutum and D. rectum after the initiation of water-stress treatments (mean ± SEM, n = 5). Midday leaf water potentials on day 21 were all below −4.0 MPa (minimum for the pressure chamber used) or leaves were dead

Full size image

Osmotic potential and adjustment

At all times osmotic potential of leaves from well-watered plants was lower than the coinciding leaf water potentials, indicating that leaf turgidity was maintained (Fig. 3). In plants subjected to water-stress, osmotic potential declined with increasing stress. In M. sativa and D. hirsutum, leaf water potentials were lower than osmotic potentials (indicating a loss of turgor) at midday after day 10 and predawn after day 17. D. rectum leaves had positive turgor in all cases, except at midday on day 21.

Concentration effects on changes in leaf osmotic potential were removed by presenting the osmotic potential at full turgor (Fig. 4). The π sat of D. hirsutum and D. rectum leaves from droughted plants was significantly lower than well-watered plants onwards from day 13 and day 17, respectively. On day 21, D. hirsutum had osmotically adjusted by 0.68 MPa and D. rectum by 0.54 MPa (Fig. 4b, c). M. sativa displayed no clear accumulation of osmotica as water-stress increased, but π sat was 0.3 MPa lower in droughted plants at day 21 (Fig. 4a). In all species, osmotic potential (Fig. 3d–f) and π sat (not presented) was lower at midday than at predawn by 0.15–0.19 MPa.

Fig. 4
figure 4

Changes in calculated osmotic potential at full turgor (π sat) in leaves of M. sativa (a), D. hirsutum (b) and D. rectum (c) sampled at midday from well-watered (solid) and water-stressed plants (open; mean ± SEM, n = 5; l.s.d. at P < 0.05)

Full size image

Photosynthetic responses

All species showed reduced photosynthesis and stomatal conductance in response to declining water status (Fig. 5). Stomatal conductance in M. sativa and D. hirsutum was substantially reduced once leaf water potential declined below −1.8 MPa. Photosynthesis in M. sativa and D. hirsutum declined in response to similar reductions in leaf water potential (Fig. 5a, b). However, stomatal conductance and photosynthesis of water-stressed D. rectum plants were reduced even though no reduction in leaf water potential or leaf turgor was observed.

Fig. 5
figure 5

Relationship between midday leaf water potential and photosynthetic rate (μmol m−2 s−1) and stomatal conductance (gS w; mmol m−2 s−1) for well-watered (solid) and water-stressed plants (open) of M. sativa (a), D. hirsutum (b) and D. rectum (c)

Full size image

Stomatal conductance and photosynthetic rate of well-watered D. rectum plants were lower than for D. hirsutum and M. sativa (Fig. 5). For example, mean stomatal conductance on a leaf area basis was 1.15 ± 0.07 mol m−2 s−1 for M. sativa and 1.00 ± 0.11 mol m−2 s−1 for D. hirsutum compared to 0.74 ± 0.06 mol m−2 s−1 for D. rectum. However, photosynthetic rate of D. hirsutum was lower than for M. sativa on a leaf dry mass basis, being more similar to D. rectum (M. sativa, 523 ± 38 μmol g−1 s−1; D. hirsutum, 346 ± 68 μmol g−1 s−1 and D. rectum, 371 ± 20 μmol g−1 s−1). Despite differences in maximum rate of photosynthesis of well-watered plants, photosynthetic WUE (A/gS w) was 33 ± 2 μmol mol−1 for D. rectum, 23 ± 2 μmol mol−1 for M. sativa and 26 ± 3 μmol mol−1 for D. hirsutum (Fig. 6). A similar ratio of internal to external CO2 concentrations was also calculated in the three species at these times (Fig. 6). WUE increased at the last measurement in all species when internal CO2 concentration declined (Fig. 6). This coincided with higher midday leaf water potentials in D. rectum (−2.6 MPa) than in M. sativa or D. hirsutum (<−4.0 MPa).

Fig. 6
figure 6

Changes in photosynthetic water use efficiency (A/gS w) and the ratio of internal and external CO2 concentration (C i/C a) with declining soil water content in M. sativa (a), D. hirsutum (b) and D. rectum (c) (mean ± SEM, n = 5; dotted line represents soil water potential of −1,500 kPa)

Full size image

Leaf morphological adaptations

A number of leaf morphological adaptations to water deficit were observed among the three species. Trichomes (leaf hairs) were present on the leaves of D. hirsutum on both the abaxial and adaxial leaf surfaces (Plate 1c, d). Leaf hairs were present in D. rectum to a lesser extent, but were not observed on M. sativa (Plate 1a, b). Both D. hirsutum and M. sativa leaves responded to reduce radiation interception as water deficit developed. M. sativa leaflets folded to form a cup, while D. hirsutum leaves folded flat against the stem and were oriented parallel to incoming solar radiation (Plate 2). Specific leaf area of droughted D. hirsutum plants was also lower than in well-watered plants after day 6, while no difference was observed for M. sativa or D. rectum except on day 21 (Fig. 7).

Plate 1
figure a

Comparison of leaf pubescence in M. sativa (a), D. rectum (b) and D. hirsutum (c). Inset (d) illustrates the presence of trichomes on abaxial and adaxial leaf surfaces of D. hirsutum

Full size image
Plate 2
figure b

Leaf movements to reduce effective leaf area and radiation interception by water-stressed plants of M. sativa (a) and D. hirsutum (b)

Full size image
Fig. 7
figure 7

Changes in specific leaf area of fully expanded leaves of M. sativa (a), D. hirsutum (b) and D. rectum (c) sampled at midday from well-watered (solid) and water-stressed plants (open; mean ± SEM, n = 5)

Full size image

Recovery after drought

Watering was withheld for an additional 2 weeks after the completion of physiological measurements, during which time SWC was below 4% w/w (i.e. soil water potential <−1.5 MPa) in M. sativa and D. hirsutum (Fig. 2b). Depletion of soil water was low during this period and leaves and stems of both species had senesced. After pots had been rewatered to field capacity for 10 days, all five D. hirsutum plants and four out of five M. sativa plants regrew. SWC was higher in D. rectum over this period and yet only one out of five D. rectum plants survived to recover upon rewatering. It appeared that once all green leaves had senesced from D. rectum this indicated that the plants had died, as none responded to rewatering after they had reached this condition.

Discussion

This study has identified some distinct physiological and morphological adaptations to water deficit among D. rectum, D. hirsutum and M. sativa. D. hirsutum, in particular, exhibited leaf physiological and morphological adaptations associated with resistance to water-stress and adaptation to arid environments. The water relations and poor survival of D. rectum under water deficit shows that this species is poorly adapted to environments where drought is common.

Physiological adaptations

In the present study, D. hirsutum and M. sativa displayed similar decline in plant water status as water deficit increased. Reductions in stomatal conductance for both species were also associated with similar midday leaf water potentials (<−1.8 MPa). In field-based studies, M. sativa stomatal conductance decreased at leaf water potentials of −1.2 to −1.5 MPa with a minimum typically reached at −2.5 MPa (Sheaffer et al. 1988). Similar stomatal responses could be expected from D. hirsutum and M. sativa grown under field conditions.

On the other hand, D. hirsutum displayed a greater degree of osmotic adjustment than M. sativa. This osmotic adjustment had little effect on maintaining leaf photosynthesis or transpiration, as it only occurred from day 13 when midday leaf water potentials were <−2.0 MPa and the associated reductions in stomatal conductance and photosynthesis had already occurred. Nonetheless, this lowering of plant water potential at full turgor may be useful to extract additional water from the soil and may play a role in conditions where water-stress develops more slowly or in subsequent wetting and drying cycles. Osmoprotectants, including proline and proline betaine, have been found to accumulate in leaves and the phloem of M. sativa subjected to water- and salt-stress (Irigoyen et al. 1992b; Girrouse et al. 1996; Trinchant et al. 2004). While the concentrations of osmoprotectants were not measured in this study, osmotica did not accumulate in response to increasing water-stress in M. sativa. This may have been a consequence of the rapid development of water-stress or genetic differences in the capacity of M. sativa cultivars to produce osmoprotectant molecules (Wood et al. 1991).

The water relations of D. rectum were somewhat puzzling and differed from those for D. hirsutum and M. sativa. Leaf water potential was similar at midday and predawn and D. rectum plants maintained a higher plant water status than the other species at the same level of available soil water. These features may be achieved by high cell elasticity, rapid translocation of water and/or water storage in plant tissues. The thick fleshy stems of D. rectum were consistent with the latter scenario, but the process by which leaf water potential is maintained is unknown. Once soil water was insufficient for plants to maintain high plant water status, leaf water potential declined quickly, suggesting that D. rectum does not have appropriate strategies to adjust to reduced water supply. D. rectum did display some osmotic adjustment, but this occurred at plant water status greater than required to reduce rates of photosynthesis and stomatal conductance. Despite the apparent maintenance of leaf water potential, D. rectum exhibited a reduction in stomatal conductance in water-stressed plants where leaf water potential had not declined. Begg and Turner (1976) assert that stomata close when leaf turgor declines below a critical value, but in D. rectum stomatal closure was more closely related to RLWC. Modifications of stomatal conductance and plant transpiration from hormonal root to shoot signalling of abscisic acid may explain stomatal closure not associated with leaf turgor (Sauter et al. 2001).

In addition to differences in plant physiological responses to water-stress, differences in stomatal conductance and photosynthetic rate among well-watered plants of each species were evident. M. sativa and D. hirsutum displayed higher stomatal conductance than D. rectum when leaf water status was high. High rates of stomatal conductance similar to those measured in this study are common in M. sativa [0.028–0.033 m s−1; or 1.25–1.47 mol m−2 s−1; converted from mol m−2 s−1 to m3 m−2 s−1 or m s−1 using molar volume of ideal gas, i.e. 0.02241 m3 mol−1] (Carter and Sheaffer 1983b; Sheaffer et al. 1988) and might be indicative of higher stomatal density and/or larger stomatal apertures than D. rectum. The differences in the photosynthetic rate of well-watered plants among the species were consistent with their plant growth rate characteristics measured previously (Bell 2005). M. sativa had a higher relative growth rate (i.e. growth increment per unit of plant mass) than D. hirsutum and D. rectum (Bell 2005), and, as anticipated, displayed a higher rate of photosynthesis per gram of leaf. The lower photosynthetic rate per unit leaf area observed for D. rectum in the present study was analogous with the lower net assimilation rate previously recorded (i.e. growth increment per unit leaf area; Bell 2005). In the present experiment and Bell (2005) assimilation per unit leaf area was similar for D. hirsutum and M. sativa, but the lower specific leaf area of D. hirsutum explains the lower assimilation per unit mass. Despite the differences in photosynthetic rate, it is surprising that no difference in photosynthetic WUE was observed among M. sativa, D. hirsutum and D. rectum.

A result of interest was that in all species reductions in stomatal conductance were not associated with a decline in the ratio of internal to external CO2 concentration in the current experiment. Internal CO2 concentration was unaffected until the final measurement (Fig. 6), when water deficits were far greater than those required to close stomata (<−3.0 MPa midday ψ leaf) and photosynthetic rate was already reduced (Fig. 5). Similar observations have previously been made in M. sativa (Irigoyen et al. 1992a) and a range of other species (Begg and Turner 1976). CO2 exchange may be still occurring due to cuticular conductance or incomplete stomatal closure. Carter and Sheaffer (1983b) recorded stomatal conductance of 0.001–0.003 m s−1 (45–125 mmol m−2 s−1) when leaf water potential had fallen below −2.5 MPa in M. sativa. In the present study, internal CO2 concentration was maintained until stomatal conductance declined below about 50 mmol m−2 s−1. The results suggest that CO2 concentration was not responsible for the decline in rate of photosynthesis. Non-stomatal reductions in photosynthesis have been attributed to water-stress impacts on the photosynthetic pathway through the inhibition of chloroplast activity (Boyer and Bowen 1970; Boyer and Potter 1973). This is consistent with the results of the present experiment where the proportional decline in photosynthesis and transpiration as water-stress developed meant that photosynthetic or WUE did not improve.

Morphological adaptations

Leaf morphological adaptations to water-limiting conditions were observed in D. hirsutum and M. sativa and may also contribute to their greater tolerance of water deficits compared with D. rectum. The leaves and stems of D. hirsutum are particularly pubescent. Leaf pubescence increases the boundary layer and thus decreases the transpiration water loss (Grammatikopoulous and Manetas 1994). Leaf trichomes can also reduce heat load by reflecting radiation and reducing the absorption of photosynthetically active radiation (Begg and Turner 1976). Radiation interception was also reduced by changes in leaf orientation in both D. hirsutum and M. sativa. The phenomenon of leaf cupping has been previously documented in M. sativa and occurs in conjunction with midday decline in leaf water potential (Travis and Reed 1983). In this study, it was also observed that specific leaf area of water-stressed D. hirsutum plants was significantly lower than in well-watered plants. This apparent plasticity in D. hirsutum may be associated with adjustments in leaf morphological as stress increases.

It was also observed that whole stems of D. hirsutum and M. sativa died after a short period of severe water deficit, but plants resprouted upon rewatering. On the other hand, in D. rectum once all green leaves had been lost, plants failed to resprout after rewatering. It is widely recognised that M. sativa plants will shed leaves during severe water-stress and assume a semi-dormant state until water supply is improved (Carter and Sheaffer 1983a; Irigoyen et al. 1992a). D. hirsutum also appears to sacrifice stems and leaves in order to maintain water status in buds and meristems near the base of the plant during severe water-stress. It is possible that xylem vessels of M. sativa and D. hirsutum cavitate more readily than D. rectum and reduce water loss by shutting down older stems (Tyree and Ewers 1991), but this has not been documented elsewhere in these species and was not investigated in this study.