The study was conducted during 2012 (“on” year), on a 14-year-old pistachio trees cv. Kerman orchard, situated in “La Entresierra” research station, Ciudad Real (Central Spain) (L 3º56’W; L 39º0’N; altitude 640 m). Pistachio trees were budded on three different rootstocks: P. terebinthus, P. atlantica and P. integerrima. Tree spacing was set at 7 m × 6 m (238 trees/ha). Pistachio trees, cv. Peter, were used as male trees and were distributed uniformly throughout the field, in a proportion of 10%. Irrigation was performed daily using a drip irrigation system with twelve self-compensating emitters (each delivering 4 L/h tree). Irrigation water has an electrical conductivity of 2.6-2.9 dS/cm. Irrigation lines were separated 1 meter from the tree line and the distance between drippers was 1 m as well (Fig. 1).

The soil at the experimental site is an alkaline (pH 8.1) shallow soil with a discontinuous petrocalcic horizon located at 0.50 m (Petrocalcic Palexeralfs), with a clay loam texture, low electrical conductivity (0.2 dS/m), 1.05% of organic matter, 0.12% of nitrogen, 17^∙·^∙10⁻⁴ mol/kg of potassium levels and high cationic exchange capacity (0.186 mol/kg). The soil volumetric water content for the first 0.3 m depth is 22.8% at field capacity (soil matric potential -0.03 MPa) and 12.1% at permanent wilting point (soil matric potential -1.5 MPa), and from 0.3 to 0.5 m it was 43.0 and 21.1%, respectively.

The climate is Mediterranean with an average annual rainfall of 397 mm, mostly distributed outside a four-month summer drought period. Rainfall, air temperature, solar radiation and humidity data were collected from an automatic weather station situated near the field.

The experimental design was factorial and completely randomized. Two factors were involved: rootstocks and irrigation; with three levels for each factor. The three rootstocks used were PT, PA and PI. These rootstocks were irrigated with three different treatments: a Control and two RDI strategies, T1 and T2.

The phenological stages taken into account in the implementation of RDI treatments were those suggested by Goldhamer (1995): stage I, from sprouting until the end of rapid nut growth (when pericarp reaches its maximum volume); stage II, from maximum nut size until the beginning of kernel growth (commonly known as the period of shell hardening); and stage III, from the beginning of kernel growth until harvest (this period starts after a fresh weight lag phase behaviour).

Contrary to stone fruits, the growth of the pericarp and the growth of the seed (kernel) in pistachio fruits are separated in time. The period separating these two processes (stage II of fruit growth) has been reported to be the most tolerant phase to water stress and make the subject of water deficit phase for water stress treatments.

Each experimental plot was formed by two trees used for measurements, surrounded by trees under the same conditions (rootstock and irrigation). In total, eighteen plots were established in the field. Six plots for each irrigation treatment and six for each rootstock. The design was factorial and completely randomized.

The three water regimes were:

Control trees: During the irrigation season, control plants received water to cover estimated crop evapotranspiration (ETc) losses by fully replenishing all soil water extracted. ETc was estimated according to daily reference evapotranspiration (ETo), calculated using the Penman–Monteith equation (Allen et al., 1998), a crop factor for pistachio (Kc) based on the time of the year (Goldahmer, 1995) and a corrector coefficient for the canopy size (Kr) (Fereres & Goldhamer, 1990). The mean Kc values used were 0.65, 1.19 and 1.10 for stages I, II and III, respectively (Goldhamer, 1995). Kr was set according to the canopy volume of each irrigation treatment, as no significant difference was found previously in the canopy volume of the different rootstocks (Memmi et al., 2015). In order to ensure the maintenance of no-water-stress conditions and to establish a reference, midday Yx and soil water balance were considered to check and modify the ETc in case of misestimating. A Ψ_x value of -1.0 MPa, considered for several woody crops as a situation of an absence of water stress (Shackel et al., 1997) was taken as threshold. When Ψ_x values of that week were below -1.0 MPa, irrigation amounts were increased a 10%, as suggested by Goldhamer & Fereres (2001).

T1 trees: This treatment was irrigated as control along the season except during stage II. During this stage, T1 trees were not maintained to the same midday Ψ_x reference indicated for control (theoretical value of -1 MPa) but have been stressed to drop their midday Ψ_x value to a threshold of -1.5 MPa.

T2 trees: This treatment was stressed during the same period than T1 but led to a higher level of stress in which midday Ψ_x threshold was maintained to a value of around -2.0 MPa during stage II.

In order to be more accurate, the implementation of RDI in this work is not governed by the traditional method of applying a percentage of the ETc applied to control treatment but tries to combine the benefits of RDI on water saving with the benefit of midday Ψ_x to detect stress. To achieve this goal, the irrigation protocol for stressed treatments was derived from the methodology proposed by Moriana et al. (2012) for olive trees and was adapted to pistachio by considering a midday Ψ_x threshold of -1.00 MPa as an index for no water stress conditions;

Irrigation started when the measured Ψ_x values were lower than the threshold values considered. The first irrigation event was always 1 mm. Then, each increase or decrease of irrigation water supply was scheduled according to the percentage of difference derived from the relation between measured Ψ_xx and the set threshold for that stage: if deviation rate was <10%, the variation in the irrigation was 0.25 mm/day; if deviation rate was between 10 and 20%, the variation in the irrigation was 0.5 mm/day; if deviation rate was between 20 and 30%, the variation in the irrigation was 1 mm/day; and if deviation rate was >30%, the variation in the irrigation was 2 mm/day. Irrigation was stopped if this approach indicated a negative application.

The monitoring of soil moisture was done using a portable capacitance probe model (Diviner 2000 Sentek Pty. Ltd., Stepney South, Australia) employing the default calibration supplied by the manufacturer. Four access tubes measuring 50 cm in length were installed for each experimental plot. Two access tubes were located in the middle between tree axis and drip line, at 50 cm from each one. One access tube was installed in front of a dripper at 50 cm from the drip line, and 50 cm from tree axis. The other access tube between drip line and tree axis was located in front of the middle of two drippers, at 50 cm from the other access tube.

The other two access tubes were located in the same perpendicular line from the tree axis as the previous access tube, in the middle, between drip line and the middle of the row, in fact, at 1.75 m from the other two access tubes, and 2.25 m from the tree axis (Fig. 1).

A weighted average based on the area of each moisture zone was calculated to determine the soil water content. Further details are presented in Memmi et al. (2015).

Real crop water consumption for each treatment and period was derived from the soil water balance calculated according to FAO methodology (Allen et al., 1998). Irrigation and effective precipitation considered as 70% of total precipitation according to the recommendations of Orgaz & Fereres (2008) for this zone and type of soils, were taken as soil water inputs while actual crop evapotranspiration (ETa) was established as soil water outputs. Soil water content balance was taken into account as input or output depending on its signs. Deep percolation of irrigation water, runoff and capillary rise were considered as minimum due to the irrigation system type and soil characteristics and were supervised by an established moisture pattern for this field.

Water relations were characterized along the season, with three diurnal cycles of midday Ψ_x and abaxial leaf conductance. Ψ_x was measured using a pressure chamber (Soil Moisture Equip., Santa Barbara, CA, USA) (Hsiao, 1990). One leaf was taken from each tree in each experimental plot for measurements. The leaf subject to measurement was selected from the middle and shaded side of the canopy to avoid a sharp change in its state. Then it was covered with aluminium foil 1 hour before measurement, stopping transpiration, and enabling the leaf to come into balance with the water potential of the stem (Begg & Turner, 1970). Owing to the difficulty to determine Ψ_x in pistachio trees due to turpentine exudation, a piece of blotting paper was placed above the petiole cutoff to facilitate measuring as paper does not become moist with turpentine, but it does with xylem water. Measurements of abaxial leaf conductance were taken from two sun-exposed and fully expanded leaves per tree from predawn to sunset using a steady state porometer (Model LICOR-1600, UK). Both parameters were measured throughout the day, approximately every two hours from predawn until dusk.

The daily cycles were performed on: 2 July 2012, corresponding to phenological stage II and a water stress period; 24 July 2012, corresponding to the beginning of change from stage II to stage III and the initiation of rehydration (one week after the start of rewatering); during this day, cloudiness was the cause of stopping measurements earlier than expected; and 7 August 2012, corresponding to stage III and a period of no water stress.

The results were subjected to ANOVA. Means were compared using the Tukey test. Significance was set at p < 0.05. Data were also subjected to regression analysis. To check the regression model hypothesis (linearity, homoscedasticity, normality and independency), Kolmogorov-Smirnov was used with the Lilliefors correction and the Shapiro-Wilk tests for normality and Bartlett’s test for homoscedasticity. Linearity was observed in the graphics and independency was assumed due to the way that data were obtained.

Slopes and F-interception were analysed comparing equality of variances first by Bartlett’s test. When homoscedasticity was checked, slopes and f-interception were compared using an F-test, where significance was set at p < 0.05.

The meteorological conditions during the daily cycles are presented in Fig. 2. The minimum temperature was respectively, 11.82 °C, 16.91 ºC and 13.51 ºC and the maximum was 31.41 ºC, 34.27 ºC and 33.67 ºC. From the first to the last day of the study, the maximum values of VPD were 4.04 kPa, 4.24 kPa and 4.42 kPa respectively and it was registered approximately in the early afternoon.

The water balance along the study is presented in Table 1. During stage I, there was no water stress in any of treatments. From the commencement of stage II until 02/07/2012, Control treatment was irrigated with 114.41 mm, which means 81% of ETc, although there was a soil water accumulation positive balance of 9 mm. T1 was irrigated only with 33.46% and T2 a 6.45% of ETc. The reduction in irrigation produced a consumption of the soil water reserve, which was reflected by a negative value of soil water balance (Table 1). In the following two weeks, irrigation in Control was 62 mm, 40 mm in T1 and 24 mm in T2, but with a soil water balance near zero, although negative in both deficit treatments, minor reductions of water consumption (based on ETa) was produced compared to Control.

When initiating the rewatering, irrigation was the same for all treatments, around 40 mm; this quantity was sufficient to cover crop water needs for all treatments without any additional depletion in soil water reserve. Water consumption was around 5 mm/day. Similar results were obtained in the period from 23/07/2012 to 06/08/2012 (Table 1).

The measurement of water potential at predawn has usually been taken as a reference to detect water stress in plants in several studies because it is assumed that the potential before dawn depicts the moment of equilibrium between tree and soil water potential (Shultz, 1996; Williams & Araujo, 2002). In this study, measurement of Ψ_x before dawn had showed its limitation even when statistical water stress differences were detected between treatments from the fact that it was not representative of a steady state; the maximum of Ψ_x did not take place only at predawn but a few hours later, when trees were in a rehydration process from stage II to stage III (Figs. 3B and 3C). In fact, the soil-plant water potential unbalance just before sunrise is quite common and has been mentioned in several studies (Sellin, 1996; Schmidhalter, 1997; Schmidhalter et al., 1998).

The fact that predawn water potential is biased by the spatial distribution of soil moisture makes it a weak recommended indicator for tree water status determination (Ameglio et al., 1999). Even when the soil humidity is uniform, Donovan et al,. (2001) found disequilibrium between predawn water potential and soil water potential for well irrigated plants and attributed a great part of the unbalance to a putative mechanism of high concentration of leaf apoplastic solutes and/or to night time transpiration. Therefore, the use of a more specific indicator related to the current status of plant could be more useful. Unlike predawn Ψ_x measurements, significant differences have been found at midday or mid-afternoon with considerably smaller coefficients of variation (data not shown). During this time interval, the detection of water stress in pistachio trees could be the most adequate. In fact, measurement at midday showed its efficiency in several deciduous trees such as peach, almond and prune (Shackel et al., 1997) due to their high sensitivity to water deficit (McCutchan & Shackel, 1992). In pistachio, at a practical level, measurement of Ψ_x has the advantage of not being limited to midday time as with many other fruit trees for a better sensing of water stress, and a measurement at mid-afternoon might be more representative of water status than midday. In fact, differences between treatments were more evident at mid-afternoon than at midday. This practical specificity in pistachio is mainly due to its particular stomatal behaviour since the lower values of Ψ_x were obtained approximately at the moment of maximum evaporative demand producing the temporary imbalance, which occurs between the variant rate of roots water absorption, the xylem water transport and the stomata water loss (Medrano & Flexas, 2003).

The use of plant water status measurements has been suggested previously as a tool for irrigation scheduling (Shackel et al., 1997; Goldhamer et al., 1999; Moriana & Fereres, 2002), given the fact that the Ψ_x measurement results in an accurate state of the dynamic relation integrated by water stress and environmental conditions (Moriana et al., 2012). The reference to a baseline under well-watered conditions and varying atmospheric situations is required for a correct use of midday/middle-afternoon Ψ_x measurements as indicators in pistachio irrigation scheduling. The fact that this baseline is not affected by the phenological stage of the tree is very useful. As a primarily approximation, the relation represented in Fig. 4A, obtained between Ψ_x and VPD could be a useful tool for this purpose. Further works are needed in order to evaluate the extrapolation of this equation to other conditions.

According to several studies of stomatal behaviour in fruit trees, it was found that stomatal response throughout the day is characteristic of each species (Acherar & Rambal, 1992). During stage II, in fully irrigated trees, no closing response of g_l was observed, reaching its peak approximately the same time VPD reached its maximum. A strong relationship was found between g_l in fully irrigated trees and VPD (R ²=0.96), indicating that there is a low stomatal closure regulation for pistachio trees when water is not a limiting factor in this stage (Fig. 7A). This correlation presented a lower fit as water stress increased (Figs. 7B and 7C), which suggests that the association between g_l and VPD is mainly governed by the water stress level in the tree. In this case, these results advocate that under water stress, tree gas exchange was more affected by factors related to water soil availability than by atmospheric components in pistachio.

During stage III, gas exchange was clearly affected in Control trees with a decrease by 20% in g_l levels between stage II and stage III, and maximum g_l peak was observed approximately 7 hours earlier (Figs. 6A and 6C). In addition, the correlation between VPD and g_l on the third day was weaker compared to the first two days of the experiment, taking into account there were no significant changes in VPD levels throughout the experiment. Therefore, this change in g_l pattern can be regarded as physiological changes affecting tree water relations from a stage to another, rather than VPD or soil moisture conditions. Several authors reported that changes in water relation parameters in pistachio during stage II is characterized by an increase in g_l values (Sajjadinia et al., 2010; Gijon et al., 2011). In almonds, several authors suggested that the seasonal decrease in gas exchange rates were related to changes in the hormonal balance or development of leaves (Sanchez-Blanco et al., 1991; Klein et al., 2001).

Depending on the phenological stage, the stomatal response of a full irrigated pistachio trees oscillates between two different behavioural “strategies”: (i) a strategy where the stomatal opening is substantial and prolonged during the day, approaching to an apple tree behaviour or (ii) a more conservative strategy which avoids high water loss traduced by a partial stomata closure at the middle of the morning which is more similar to almonds or olives trees behaviour. The first strategy is observed during shell hardening, a process of high energy consumption which could be the cause of little noticeable water stress effect at the final yield according to Goldhamer & Beede (2004). The second strategy, a more conservative approach, is recorded during the kernel filling stage which is considered imperative to the survival of the species.

On the other hand, water stress reduced the maximum g_l in deficit treatments (T1 and T2). Under these conditions, the g_l pattern followed by T1 and T2 treatments showed the ability of pistachio trees to regulate stomatal activity under water stress conditions by closing stomata early in the morning, anticipating an excessive decline in water potentials. This control from roots to shoots could be assigned to a hydraulic or/and a biochemical signal, which prevents excessive water loss as in other fruit trees: orange, Fereres et al. (1979); almond, Castel & Fereres (1982), Torrecillas et al. (1988); olive: Bongi & Loreto (1989), Giorio et al. (1999).

Three weeks after the beginning of rehydration, a maintained positive soil water balance did not allow stressed trees to recover completely from stress levels of -1.5 MPa and -2.0 MPa maintained approximately for 40 days. Torrecillas et al. (1996) mentioned that a rapid recovery after a stress condition can be related to a greater physiological tolerance to drought, so the delay observed in the recovery of pistachio trees was not expected due to the great reputation of pistachio as a drought-tolerant species. Hence, the faster or slower response to water renewal may be related to the extent of drought (level of stress and duration) and its possible progressive effects on different plant organs (short and long term effects).

Guerrero et al. (2005) imposed a similar extent of water stress as T1 and observed a rapid recovery for g_l in pistachio trees; however, they concluded that more than three weeks are required for a complete recovery of water status in pistachio because of lagging trunk growth. Moreover, Rieger (1995) mentioned for potted pistachio plants the possibility that stomatal conductance could recover before hydraulic conductivity. In this study, a partial recovery was observed 3 weeks after the beginning of rewatering but a total recovery could not be confirmed when Ψ_x drops to values around -2.2 MPa during stress period (T2 treatment).

Several studies involving acoustic sensing of woody tissues in other species showed that cavitation often occurred when the water potential falls below -1.5 MPa (Dixon et al., 1984; Peña & Grace, 1986; Tyree & Sperry, 1989). Then the recovery delay observed in this study may also be related to the complexity of the process for the refilling of xylem cavitation. The variant time necessary for each individual mechanism involved in the recovery to come to pre-stress conditions translates the complexity of the recovery process initiated by re-watering.

The unexpected delay for the recovery of pistachio from water stress should be taken into account and exceeding rewatering should start early for RDI major benefits specially under shallow soil conditions combined with drip irrigation system.

The three different rootstocks studied led to different water relations in Kerman cultivar. During stage II, PA showed the lowest Ψ_x and g_l values during the day (Figs. 5 and 8). Spiegel-Roy et al. (1977) and Lin et al. (1984) described PA as an adapted species for very severe conditions and dry areas where higher stomatal control is exerted by PA compared to PI and PT. PT compared to other Mediterranean plants was considered by Sakcali & Ozturk (2004) as a water spender and described it as a partially water-tolerant species which cannot sustain long in harsh conditions. These results differ from those reported by Germana (1997) and Gijón et al. (2010) for potted pistachio plants, where PA is considered as a rootstock with little stomatal control compared to PT. In this study, PA maintained lower g_l than PT and PI during stage II, but during stage III, PA displayed a greater g_l than PT early morning and similar activity during the rest of the day. This variance between results could be explained by the wide genetic diversity that affects the spectrum of potential responses, and by the different responses given by the rootstock depending on whether it is combined or not to a scion.

Ψ_x of trees budded on PI was significantly lower than that of the two other rootstocks from the second diurnal cycle. Despite this, PI presented higher g_l than the other two rootstocks. This result suggests that PI has less stomatal control than PT and PA and no discernible threshold for water potential maintenance, which could give it a little anhysohidric tendency. Duhme & Hinckley (1992) reported that the values of leaf osmotic potential at full turgor for PI were only 0.5 to 1.0 MPa higher than osmotic potentials at the turgor loss point for PT. This confirms that PI is less adapted than the other rootstocks to conditions of water scarcity. PI has been described as the most trunk and canopy size vigorous rootstock amongst the three (Ferguson et al., 2005), which could provide a higher capacity for energy interception and water transport, and can result in a major loss of water. According to the general trend of g_l, the influence of rootstocks on the degree of leaf conductance control when they are combined with cv. Kerman could be classified in this order, from the greatest to the lowest: PA > PT > PI.

In summary, midday stem water potential could be considered as a better indicator for water stress in pistachio than predawn stem water potential. Water stress conditions modified significantly the fit of the relationship between stem water potential and VPD. However, the phenological stage has no effect on this relationship. Relationship between leaf conductance and VPD has shown a positive linear correlation. Such behaviour was recorded during stage II of fruit growth where pistachio follows the tendency of a water spender species when all water requirements are fulfilled. Once water stress took place, pistachio showed an ability to reduce its water consumption by closing the stomata early in the morning. The recovery from water stress levels reaching -1.5 MPa and -2.0 MPa prolonged for 40 days could be considered a slow process since it takes around 4 weeks. Therefore, it should be taken into account for deficit irrigation scheduling. Concerning rootstocks, in contrast with previous works and to the general belief, Pistacia atlantica showed it could be a suitable rootstock for water deficit conditions. While, Pistacia integerrima showed less stomatal regulation, promoting water loss despite the major drop of water potential and showed less adaptation than Pistacia atlantica and Pistacia terebinthus to water stress conditions.