If soil water-based measures are to be replaced by plant-based measures it is important to consider what measures might be most appropriate for irrigation scheduling purposes. Possible measures include direct measurements of some aspect of plant water status as well as measurements of a number of plant processes that are known to respond sensitively to water deficits. One might expect that a direct measure of plant water status should be the most rigorous and hence the most useful indicator of irrigation requirement, although the question remains as to where in the plant that quantity should be measured. In practice, as has been argued strongly by Jones (1990b), most plants exercise some measure of autonomous control over their shoot or leaf water status, tending to minimize changes in shoot water status as the soil dries or as evaporative demand increases (Bates and Hall, 1981; Jones, 1983). In the long term, this control is achieved through changes in leaf area and root extension, and in the shorter term through changes in leaf angle, stomatal conductance, and hydraulic properties of the transport system. In extreme cases, plants with good endogenous control systems maintain a stable leaf water status over a wide range of evaporative demand or soil water supplies; these plants are termed ‘isohydric’ (Stocker, 1956), and include especially plants such as cowpea, maize, and poplar (Bates and Hall, 1981; Tardieu and Simonneau, 1998). This is by contrast with those species such as sunflower or barley which appear to have less effective control of leaf water status and have been termed ‘anisohydric’. In practice the distinctions between isohydric and anisohydric behaviour are often not clear-cut; even different cultivars of grapevine have been shown to have contrasting hydraulic behaviours (Schultz, 2003).
The choice of which plant-based measure to use depends on their relative sensitivity to water deficits. The definition of sensitivity, however, is somewhat problematic. The relative sensitivities of different physiological processes were reviewed in some detail by Hsiao (1973), who identified cell growth as being most sensitive to tissue water deficits, closely followed by wall and protein synthesis, all of which could respond to water deficits of less than 0.1 MPa (Fig. 1). Hsiao reported that stomatal closure was only rarely affected when tissue water potential fell by 0.2–0.5 MPa, with decreases of 1.0 MPa or more being required for stomatal closure in many cases. Although photosynthesis was classified as moderately sensitive by Hsiao, largely as a result of its dependence on stomatal aperture, some component processes such as electron transport are now known to be particularly insensitive (Massacci and Jones, 1990). It is now believed that Hsiao's (1973) classification is somewhat misleading, and underestimates the true sensitivity of the stomata, as it is based on observed responses to leaf water potential alone and ignores the internal root–shoot signalling that is now known to play a major part in controlling stomatal aperture (Davies and Zhang, 1991).
The error arising from a reliance on leaf water status is readily apparent when one considers that many plants operate when optimally watered with the leaf water potential at around –2 MPa, yet the stomata may close as the soil dries by only a few tens of Pa, with little change in leaf water potential (Bates and Hall, 1981
). A further consideration is that any attempt to relate stomatal aperture to leaf water potential in a long-term drought experiment can also be misleading, because with slowly developing stress the plant adapts by decreasing leaf area; as a result stomatal conductance and photosynthesis rate per unit leaf area may remain fairly stable as soil dries (Moriana and Fereres, 2002
). Nevertheless, over shorter time-scales it still appears that stomata are a particularly sensitive early indicator of water deficits.
In principle, water status is not ideal as a measure of water deficit as it is already subject to some physiological control, and indeed, as has been outlined above, leaf water potential generally shows some homeostasis. Nevertheless, changes in water status somewhere in the plant system are assumed to be a prerequisite for any physiological adaptation or other response. All that a homeostatic system can do is to minimize, not eliminate, the changes in water status; indeed for a feedback system of stomatal control it is not theoretically possible for such a system to stably eliminate changes in shoot water status if that is the variable that actually controls the stomata (Jones, 1990b; Franks et al., 1997).
In general, the use of any plant-based or similar indicator for irrigation scheduling requires the definition of reference or threshold values, beyond which irrigation is necessary (Table 1). Such reference values are commonly determined for plants growing under non-limiting soil water supply (Fereres and Goldhamer, 2003), but obtaining extensive information on the behaviour of these reference values as environmental conditions change is an important stage in the development and validation of such methods. Another general limitation to plant-based methods is that they do not usually give information on ‘how much’ irrigation to apply at any time, only whether or not irrigation is needed.
Plant water status
Perhaps the first approach to the use of the plant itself as an indicator of irrigation need, and one that is still frequently adopted today, was to base irrigation on visible wilting. Unfortunately, by the time wilting is apparent a substantial proportion of potential yield may already have been lost (Slatyer, 1967). More rigorous and more sensitive measures of plant water status are therefore required. Although relative water content (RWC) (Barrs, 1968) is a widely used measure of water status that does not require sophisticated equipment, it is often argued that water potential, especially of the leaves (leaf) is a more rigorous and more generally applicable measure of plant water status (Slatyer, 1967; Jones, 1990b). In spite of this, RWC has the advantage that it can be more closely related to cell turgor, which is the process directly driving cell expansion, than it is to the total water potential (Jones, 1990b).
The fact that plant water status, and especially leaf water status, is usually controlled to some extent by means of stomatal closure or other regulatory mechanisms, argues against the use of such measures, especially in strongly isohydric species. A further problem with the use of leaf water status as an indicator of irrigation need was pointed out by Jones (1990b), who noted that even though there was often homeostasis of leaf water potential between different soil moisture regimes, rapid temporal fluctuations are often observed as a function of environmental conditions (such as passing clouds). This makes the interpretation of leaf water potential as an indicator of irrigation-need doubly unsatisfactory. Nevertheless, in spite of the concerns with the use of leaf water status that have been outlined above, it has been reported that leaf water potential can, when corrected for diurnal and environmental variation, provide a sensitive index for irrigation control (Peretz et al., 1984).
As a partial solution to the variability of leaf water status, various workers have proposed that a more useful and more robust indicator of water status is the xylem water potential or stem water potential (SWP, measured by using a pressure chamber on leaves enclosed in darkened plastic bags for some time before measurement and allowed to equilibrate with the xylem water potential; McCutchan and Shackel, 1992). As a more stable measure of water status, others have even recommended that measurements should be made on pre-equilibrated leaves from root suckers (Jones, 1990a; Simonneau and Habib, 1991). These methods are thought to be preferable largely because they approach more closely the soil water status than does the value of leaf water potential, although as a result they therefore miss out on the potential advantages of plant-based methods.
Perhaps an even better estimator of the soil water potential is the predawn leaf water potential (as leaf should largely equilibrate with soil by dawn). Unfortunately this is often found to be rather insensitive to variation in soil moisture content (Garnier and Berger, 1987). Further, this is not very convenient for irrigation scheduling as routine measurements predawn are expensive to obtain, and at best can only be obtained daily. As yet another alternative, Jones (1983) suggested the indirect estimation of an effective soil water potential at the root surface of transpiring plants based on measurements of leaf water potential and stomatal conductance during the day, and argued that this should have significant advantages over predawn measurements. Such an approach has been successfully tested by Lorenzo-Minguez et al. (1985).
None of the above plant-based methods are well adapted for automation of irrigation scheduling or control because of the difficulties of measurement of any of the variables discussed. Although it may be possible to use automated stem or leaf psychrometers (Dixon and Tyree, 1984), these instruments are notoriously unreliable. In conclusion, it is apparent from the above discussion that the favoured way to use plant water status is actually as an indicator of soil water status; this negates many of the advantages of selecting a plant-based measure! Indeed soil water potential can be measured directly, thus avoiding the need for any plant-based measurement, although it is worth noting that this does not necessarily give a good measure of the effective water potential at the root surface during active transpiration (Jones, 1983).
Several indirect methods for measuring or monitoring water status have been developed as alternatives to direct measurement. The general behaviour of a number of such methods have been compared by McBurney (1992) and Sellés and Berger (1990). In general, these indirect methods suffer from the same disadvantages as do the direct measurements of leaf water status, but in certain circumstances have been developed into commercial systems. Some of these approaches are reviewed below.
A number of instruments are available for the routine monitoring of leaf thickness, which is known to decrease as turgidity decreases. Approaches include direct measurement using linear displacement transducers (e.g. LVDTs [Burquez, 1987; Malone, 1993] or capacitance sensors [McBurney, 1992]) or through measurements of leaf ‘superficial density’ using ß-ray attenuation (Jones, 1973). Unfortunately, leaf thickness is frequently even less sensitive to changes in water status than is leaf water content because, especially with younger leaves, a fraction of leaf shrinkage is often in the plane of the leaves rather than in the direction of the sensor (Jones, 1973).
Stem and fruit diameter:
Stem and fruit diameters fluctuate diurnally in response to changes in water content, and so suffer from many of the same disadvantages as other water status measures. Nevertheless, the diurnal dynamics of changes in diameter, especially of fruits, have been used to derive rather more sensitive indicators of irrigation need, where the magnitude of daily shrinkage has been used to indicate water status, and comparisons of diameters at the same time on succeeding days give a measure of growth rate (Huguet et al., 1992; Li and Huguet, 1990; Jones, 1985). Although changes in growth rate provide a particularly sensitive measure of plant water stress, such daily measurements are not particularly useful for the control of high-frequency irrigation systems. Nevertheless, several workers have achieved promising results for low-frequency irrigation scheduling by the use of maximum daily shrinkage (MDS). For example, Fereres and Goldhamer (2003) showed that MDS was a more promising approach for automated irrigation scheduling than was the use of stem water potential for almond trees, while differences in maximum trunk diameter were also found to be particularly useful in olive (Moriana and Fereres, 2002). The use of such dendrometry or micromorphometric techniques has been developed into a number of successful commercial irrigation scheduling systems (e.g. ‘Pepista 4000’, Delta International, Montfavet, France); these are usually applied to the study of stem diameter changes. Sellés and Berger (1990) reported that variations in trunk diameter or stem water potential were more sensitive as indicators of irrigation need than was the variation in fruit diameter. This was probably a result of the poor hydraulic connection between fruit tissue and the conducting xylem. There is currently much interest in evaluating such techniques for irrigation scheduling, with a number of relevant papers presented at recent meetings (e.g. the International Society for Horticultural Science 4th International Symposium on Irrigation of Horticultural Crops, 1–5 September 2003, Davis, CA, USA [as yet unpublished], and Kang et al., 2003).
A related approach to the study of changes in stem water content was the use of -ray attenuation (Brough et al., 1986). Although this was shown to be very sensitive, safety considerations and cost have largely limited the further application of this approach.
The development of reliable heat pulse and energy balance thermal sensors for sap-flow measurement in the stems of plants (Granier, 1987; Cohen et al., 1981; Cermak and Kucera, 1981) has opened up an alternative approach to irrigation scheduling based on measurements of sap-flow rates. Because sap-flow rates are expected to be sensitive to water deficits and especially to stomatal closure, many workers have tested the use of sap-flow measurement for irrigation scheduling and control in a diverse range of crops, including grapevine (Eastham and Gray, 1998; Ginestar et al., 1998a, b), fruit and olive trees (Ameglio et al., 1998; Fernandez et al., 2001; Giorio and Giorio, 2003; Remorini and Massai, 2003) and even greenhouse crops (Ehret et al., 2001).
Although the changes in transpiration rate that sap flow indicates are largely determined by changes in stomatal aperture, transpiration is also influenced by other environmental conditions such as humidity. Therefore changes in sap flow can occur without changes in stomatal opening. Even though rates of sap flow may vary markedly between trees as a result of differences in tree size and exposure, the general patterns of change in response to both environmental conditions and to water status are similar (Eastham and Gray, 1998). Appropriate sap-flow rates to use as ‘control thresholds’ may be derived by means of regular calibration measurements, especially for larger trees. Alternatively, it is at least feasible in principle to derive an irrigation scheduling algorithm that is based on an analysis of the diurnal patterns of sap flow, with midday reductions being indicative of developing water deficits (though of course diurnal fluctuations in environmental conditions can mimic such changes). Another potential problem with sap flow for precision control is that it tends to lag behind changes in transpiration rate owing to the hydraulic capacitance of the stem and other plant tissues (Wronski et al., 1985).
It follows that, although sap-flow measurement is well adapted for automated recording and hence potentially automated control of irrigation systems, it can be a little difficult to determine the correct control points for any crop.
It is generally accepted (Steudle, 2001) that water in the xylem vessels of transpiring plants is under tension; as water deficits increase, this tension is thought to increase to such an extent that the water columns can fracture, or ‘cavitate’. Such cavitation events lead to the explosive formation of a bubble, initially containing water vapour. These cavitation events can be detected acoustically in the audio- (Milburn, 1979) or ultrasonic-frequencies (Tyree and Dixon, 1983), and the resulting embolisms may restrict water flow through the stem. Substantial evidence, though largely circumstantial, now indicates that the ultrasonic acoustic emissions (AEs) detected as plants become stressed do indeed indicate cavitation events and that AE rates can be used as an indicator of plant ‘stress’ (Tyree and Sperry, 1989). Nevertheless, there remain many uncertainties as it seems that at least a proportion of the AEs detected as woody tissues dry out may not be related to xylem embolisms. For example, the large numbers observed by Sandford and Grace (1985) as coniferous stems dried out were substantially in excess of the number of conducting tracheids present, thus suggesting a major contribution to observed AEs by the non-conducting fibres (Jones and Peña, 1986). Although the measurement of AEs has proved to be a powerful tool for the study of hydraulic architecture in plants, there has been little progress in adapting this measure as an indicator for irrigation scheduling. Note, however, the recent report by Yang et al. (2003), who implemented a control algorithm based on the association of AE rate with transpiration rate for the precision irrigation of tomato. It is likely that the main reasons for the lack of uptake include the fact that the relationship between the number of AEs and water status changes with successive cycles of stress, and the fact that cavitation events are mostly observed during the drying phase, not during rewetting, and so cannot provide an indicator of when irrigation has been sufficient to replenish the soil water supply.
Stomatal conductance and thermal sensing
As outlined above, it appears that changes in stomatal conductance are particularly sensitive to developing water deficits in many plants and therefore potentially provide a good indicator of irrigation need in many species. It is in this area that most effort has been concentrated on the development of practical, plant-based irrigation scheduling approaches. Although stomatal conductance can be measured accurately using widely available diffusion porometers, measurements are labour-intensive and unsuitable for automation. The recognition that leaf temperature tends to increase as plants are droughted and stomata close (Raschke, 1960) led to a major effort in the 1970s and 1980s to develop thermal sensing methods, based on the newly developed infrared thermometers, for the detection of plant stress (see reviews by Jackson, 1982; Jones and Leinonen, 2003; Jones, 2004).
An early method of accounting for the rapid short-term variation in leaf temperature as radiation and wind speed vary in the field was to refer leaf temperatures to air temperature and to integrate these differences (e.g. the Stress Degree Day measure of Jackson et al., 1977); significant elevation of canopy temperature above air temperature was indicative of stomatal closure and water deficit stress. The method was transformed into a more practical approach following the introduction of the crop water stress index (CWSI) by Idso and colleagues (Idso et al., 1981; Jackson et al., 1981), where CWSI was obtained from the canopy temperature (Tcanopy) according to
is a so-called non-water-stressed baseline temperature for the crop in question at the same atmospheric vapour pressure deficit, and Tdry
is an independently derived temperature of a non-transpiring reference crop (Fig. 2
). In this approach all temperatures are expressed as differences from air temperature so that standard relationships for Tdry
can be used. Although this approach was found to be useful in the clear arid climate of Arizona where the method was developed, it has proved to be less useful in more humid or cloudy climates where the signal-to-noise ratio is somewhat smaller (see Fig. 2
; Hipps et al.
; Jones, 1999
). In spite of its deficiencies, there has been widespread use of infrared thermometry as a tool in irrigation scheduling in many, especially arid, situations (Jackson, 1982
; Stockle and Dugas, 1992
; Martin et al.
), especially with the development of ‘trapezoidal’ methods involving the combination of temperature data with a visible/near infrared vegetation index (Moran et al.
In order to improve the precision of the approach in more humid or low-radiation environments, Jones (1999)
introduced the approach of using physical dry and wet reference surfaces to replace the notional Tdry
required for equation 1. A number of recent papers have shown that this approach can give reliable and sensitive indications of stomatal closure (Diaz-Espejo and Verhoef, 2002
; Jones et al.
; Leinonen and Jones, 2004
) and hence has the potential to be used for irrigation scheduling. The most important recent advances in the application of thermal sensing for plant ‘stress’ detection and irrigation scheduling, however, have been provided by the introduction of thermal imagery (Jones, 1999
; Jones et al.
), although their expense has meant that such systems have yet to be widely used.
In addition to the use of the absolute temperature rise as stomata close, it has also been proposed that use may be made of the fact that the variance of leaf temperature increases as stomata close (Fuchs, 1990). Indeed, this may be a more sensitive indicator of stomatal closure than is the temperature rise (Jones, 2004). Again, the introduction of thermal cameras now makes the wider use of such approaches feasible, especially when combined with automated image analysis.