Limitations to adopting regulated deficit irrigation in stone fruit orchards: a case study

Fruit production development is resulting in large commercial orchards with improved water management standards. While the agronomic and economic benefits of regulated deficit irrigation (RDI) have long been established, the local variability in soils and climate and the irrigation system design limits its practical applications. This paper uses a case study approach (a 225 ha stone fruit orchard) to unveil limitations derived from environmental spatial variability and irrigation performance. The spatial variability of soil physical parameters and meteorology in the orchard was characterized, and its implication on crop water requirements was established. Irrigation depths applied during 20042009 were analysed and compared with crop water requirements under standard and RDI strategies. Plant water status was also measured during two irrigation seasons using stem water potential measurements. On-farm wind speed variability amounted to 55%, representing differences of 17% in reference evapotranspiration. During the study seasons, irrigation scheduling evolved towards deficit irrigation; however, the specific traits of RDI in stone fruits were not implemented. RDI implementation was limited by: 1) poor correspondence between environmental variability and irrigation system design; 2) insufficient information on RDI crop water requirements and its on-farm spatial variability within the farm; and 3) low control of the water distribution network. Additional key words: drip irrigation evaluation; irrigation requirements; stem water potential; wind spatial variability. * Corresponding author: v.zapata@csic.es Received: 22-03-12. Accepted: 22-03-13. Abbreviations used: DUlqGlobal (global low quarter distribution uniformity); DUlqOther (low quarter distribution uniformity due to other causes); DUlqΔp (low quarter distribution uniformity due to differences in emitter pressure); EMP (early maturing peach); ETc (crop evapotranspiration, mm); ETo (reference crop evapotranspiration, mm); FI to FIV [fruit growth stages for peach (Naor, 2006)]; GIR (gross irrigation requirements, mm); K (parameter); Ka (parameter for the estimation of DUlqGlobal); Kc (crop coefficient); KrRDI (reduction coefficient for regulated deficit irrigation); MAE (mean absolute error); LMP (late maturing peach); MIT (measured irrigation time, h); MMP (medium maturing peach); n (number of emitters per tree); NIR (net irrigation requirements, mm); p (operating pressure, kPa); q (emitter flow rate, L h); RAW (readily available water); RDI (regulated deficit irrigation); RMSE (root mean square error); SID (scheduled irrigation depth, mm); SIT (scheduled irrigation time, h); Std (standard irrigation management); SWP (stem water potential, MPa); VCIA (variation coefficient intra-irrigations); VCIE (variation coefficient inter-irrigations); x (exponent). Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) Spanish Journal of Agricultural Research 2013 11(2), 529-546 Available online at www.inia.es/sjar ISSN: 1695-971-X http://dx.doi.org/10.5424/sjar/2013112-2902 eISSN: 2171-9292

orchards (Chalmers et al., 1981;Mitchell et al., 1986). This advanced technique is based on the fact that plant sensitivity to water stress varies between phenological stages and that water stress at specif ic periods of vegetative growth can help control growth and vegetative-fruit competition (Chalmers et al., 1981;Mitchell & Chalmers, 1982;Cameron et al., 2006). Fereres & Soriano (2007) reported that RDI has enjoyed more success in tree crops and vines than in field crops.
Practical protocols are required for farmers to apply such advanced irrigation techniques. Farmers require answers to issues such us: what are the standard crop water requirements, when to apply deficit irrigation, and how much deficit to apply in comparison to the standard requirements. A number of references have reported on deficit irrigation experiments in orchards. The large variability in soil, meteorology, crop species, varieties and irrigation treatments of the reported scientific work, difficult the elaboration of practical procedures for implementing RDI in commercial orchards.
The establishment of commonly accepted local, standard crop water requirements is the basis for the application of deficit irrigation techniques. The FAO-56 method (Allen et al., 1998), despite its maturity, greatly benefits from experimental testing and adjustment. Experimentation leading to the determination of local crop coefficients is very much needed in regions specialized in fruit production, since Kc recommendations are often site and year specific, and have been reported to depend on local reference crop evapotranspiration (ETo) rates, rainfall, and crop management practices (Snyder et al., 2000). In addition, research leading to the determination of the local duration of the different periods and their adjustment through different growing seasons is much needed.
Regarding irrigated agriculture, wind speed plays a relevant role in the estimation of reference evapotranspiration and crop water requirements, especially in windy areas such as the Ebro Valley (Martínez-Cob & Tejero, 2004;Martínez-Cob et al., 2010). Wind spatiotemporal variability is much higher than that of other meteors of agricultural interest, such as air temperature and relative humidity (Martínez-Cob et al., 2010). The Spanish network of agrometeorological stations (SIAR, www.magrama.gob.es/siar/informacion.asp) covers most of the irrigated areas in Spain and provides records that are assumed to be representative of a certain homogenous area surrounding the station. When promoting advance irrigation techniques, especially in windy areas, the representativeness of the meteorological available data should be analyzed.
The stress-tolerant phases in peach have been identified as stage II, the lag phase of fruit growth (Chalmers et al., 1981), and postharvest (Johnson et al., 1992). Several authors have proposed different levels of deficit irrigation depending on soil characteristics  and different periods of deficit depending on peach cycles (Goldhamer et al., 2002;Gelly et al., 2004;Dichio et al., 2007; for early, medium and late maturing peaches, respectively). Girona et al. (2003) reported that water deficit during postharvest for peach orchards should be carefully managed to avoid reduction in bloom and fruit loads during the subsequent growing season. The sustainability of RDI imposes limits to the magnitude of water stress.
Antunez- Barria (2006) reported that sweet cherries showed tolerance to mild water stress. As a consequence, these authors saw potential to adopt deficit irrigation strategies to commercial production systems. However, research has been scarce on the deficit level that should be applied. For instance, Marsal et al. (2009) suggested reducing 50% of crop water requirements at postharvest. Papenfuss & Black (2010) applied different levels of deficit from pit hardening to harvest on tart cherries. These authors reported that the concentration of soluble solids and the color (chroma) of intact fruit increased with the severity of the irrigation deficit and were inversely correlated with fruit water content.
In apricot, in the absence of dwarfing rootstocks, several management practices aiming at controlling vegetative growth have been analyzed, with RDI resulting particularly successful (Arzadi et al., 2000). Torrecillas et al. (2000) studied the effect of deficit irrigation during different apricot phenological stages (stages I, II and postharvest). These authors stated that withholding irrigation during fruit growth periods (stages I and II) decreased the fruit growth rate. However, when irrigation was restored a compensatory fruit growth rate was observed.
The success of RDI techniques strongly depends on the appropriate use of on-farm irrigation systems. In the last two decades, drip irrigation has become very popular in fruit productive areas. Irrigation evaluation procedures for drip systems have evolved following technological developments. Merriam et al. (1973) developed one of the first field evaluation manuals for drip irrigation systems. The Irrigation Training and Research Centre (ITCR, San Luis Obispo, CA, USA) standardized procedures for drip irrigation system eva-luation (Burt, 2004), defining uniformity components for a complete irrigation system. Salvador et al. (2011), analyzing crop water requirements and its application in the Ebro valley of Spain, reported that deficit irrigation (not necessarily RDI) is becoming common in the local peach and cherry orchards. There are not many references in the literature quantifying the overtaking of these irrigation practices in commercial orchards, on how farmers are adopting advanced irrigation techniques and on which are the limitations to implement them. The research reported in this paper was supported by a 225 ha case study stone fruit orchard located in the Ebro valley. The orchard had high management standards, and counted on a professional, full-time irrigation manager. The general objective of this paper was to establish the limitations imposed by environmental variability and irrigation performance on the adoption of RDI at the orchard level. Specific objectives were: 1) to assess the variability in soil water retention; 2) to characterize the small-scale spatial variability of meteorological variables affecting reference evapotranspiration; 3) to assess irrigation performance at the different cropping zones of the study orchard; 4) to characterize the theoretical irrigation water requirements for the different crops of the farm, under standard and RDI strategies, accounting with the literature and the environmental variability of the study case; and 5) to compare them with the applied irrigation depths.

The case study
The case study for this research was La Herradura stone fruit orchard (225 ha), located in Caspe (Zaragoza, north eastern Spain). Geographical coordinates are 41°17' N latitude and 0°00' 08" E longitude. This orchard was selected as representative of high management standards, and can be illustrative of modern fruit orchards in an international context. The orchard is located next to a meander of the Ebro River, flooded by the Mequinenza dam. The orchard topography is quite rough, with elevation ranging from 120 to 200 m above mean sea level. The crops at the orchard included cherry (Prunus avium, 44 ha), apricot (Prunus armeniaca, 29 ha) and peach (Prunus persica, 154 ha). Peach trees were divided in early maturing peach (EMP, 51 ha), medium maturing peach (MMP, 51 ha) and late maturing peach (LMP, 52 ha). Fig. 1 presents the crop distribution in La Herradura farm.
The irrigation system is divided in 88 drip irrigated areas (irrigation subunits), each of them commanded by an automated valve: 8 for cherry, 18 for apricot, 16 for the EMP, 22 for MMP and 24 for LMP. Each irrigated area was planted to the same fruit species (and cycle, in the case of peach). However, a number of cultivars could be planted in the same area, resulting from pollination, labor demand and commercial requirements. As a consequence, a total of 55 fruit tree cultivars were cultivated at the orchard.
According to the irrigation manager, water availability was not a problem, and the pumping cost was not relevant within the general farming costs. The manager declared to implement deficit irrigation techniques to control vegetative growth and to improve fruit quality. The manager uses the design discharge of each valve to schedule the irrigation time of each irrigated area and declared to use evapotranspiration information provided by the regional irrigation advisory service to devise a weekly irrigation schedule. Only at the general pipe of the pumping stations a water metering device was installed. This measuring device was used by the irrigation manager to test the discharge of the set of valves under simultaneous operation.
Limitations to adopting regulated deficit irrigation in stone fruit orchards: a case study 531 Figure 1. Map of the different fruit tree cropping zones (apricot, cherry, early maturing peach, medium maturing peach and late maturing peach). The location of the two landmark points H1 and H2 and the wind measurement points M1 to M13 is also presented. At each irrigated area, the irrigation layout was as follows: manifolds supplied water from the valve to the irrigation laterals. Laterals branched from the manifold on one side or -more commonly-on both sides. Two laterals were used to irrigate each line of trees, one lateral at each side of the line. Turbulent (non pressure compensating) emitters were used at the orchard, with a design discharge of 4 L h -1 . Emitters were extruded in the polyethylene irrigation laterals at 1 m intervals. As a consequence of the different plating spacing, the number of emitters per tree was variable (from 5 to 9).

Soil physical properties
A non-structured soil sampling campaign was performed at the orchard. A total of 88 soil samples (0.28 samples ha -1 ) were manually drilled until a soil depth (d) of 1.2 m if possible. The coordinates of the sampling point and the sampling depth were systematically recorded using a handheld GPS. A subsample was weighted, milled and passed through a 2 mm sieve. Volumetric stoniness (S, %) was determined for each sample from the weigh of fractions above and below 2 mm, soil bulk density (ρ b , Mg m -3 ) and the stone density. Soil bulk density and stone density were estimated as 1.40 and 2.65 Mg m -3 , respectively, in agreement with previous works in the area (Playán et al., 2000). Gravimetric f ield capacity (FC, %) and wilting point (WP, %) were estimated at the laboratory using pressure plates (Hanks, 1992). Pressures of 0.03 and 1.5 MPa were considered representative of FC and WP, respectively. The readily available water (RAW, mm) was determined as: 3 ρ w 100 100 The analysis of stoniness revealed very large differences among samples. These differences corresponded to different soil units, whose limits could be easily recognized in the field. A thematic map for soil stoniness was produced based on GPS recorded observations of soil unit discrete limits. These limits often corresponded to the changes in strata responding to changes in soil surface elevation.

Wind speed and reference evapotranspiration
Wind speed variability within the orchard was characterized using three types of weather stations: 1) the reference station (SIAR network), located at about 10 km northwest of the orchard, and considered representative of the area at a sub-regional scale; 2) two landmark stations (H1 and H2), considered representative at a local scale; and 3) the movable station, which was installed for short periods of time at 13 different locations within the orchard (called M1 to M13, Fig. 1). This station was used for a detailed study of wind variability, since wind speed was identified in this work as the principal source of spatial variability of reference evapotranspiration within the orchard. A preliminary analysis of the time variability of wind speed and direction in the area was performed using the data series available at the SIAR reference station (from January 2004 to December 2009).
The two landmark weather stations were installed at two spots with maximum and minimum elevation within the orchard, with the goal of characterizing extreme local meteorological variability (Fig. 1). These two stations, H1and H2, were installed at an altitude of 175 m and 125 m above mean sea level, respectively. Both stations recorded data from August 2007 to September 2009, with the exception of the period May to September 2008, during which station H2 was used as a movable station (see below). Spot H1 was more exposed to wind than spot H2.
Both landmark stations were equipped with a data logger and sensors to measure air temperature, air relative humidity, incoming solar radiation and wind speed and direction. Precipitation was only recorded at H1. The brand and models of these sensors were the same as those in the SIAR reference station (www. magrama.gob.es/siar/descripcion.asp). Only the wind measurement sensor differed: a propeller-type anemometer was used at the SIAR reference station (Young's Model 05103, Campbell scientific, Inc., Shepshed, Leicestershire, UK), while 3-cup-rotor anemometers (model A100R, Vector InstrumentsTM, Rhyl, UK) were installed at the landmark weather stations (H1 and H2). Sensors were monitored every 10 s, half-hour and daily averages were stored in the data logger memory for further analysis. The meteorological variables recorded at both landmark stations for the periods August 2007 to April 2008 and September 2008 to September 2009 were compared using simple linear regression analyses. The statistics used to assess the regressions were the coefficient of determination (R 2 ), the root mean square error (RMSE, data not presented) and the mean average error (MAE, data not presented). Statistical significance was assessed at the 0.05 probability level. Among the recorded meteorological variables, only wind speed showed significant, relevant differences between H1 and H2. As a consequence, further efforts were made to characterize its spatial variability at the orchard.
During the period of May to September 2008, station H2 was used as the movable station. This station was moved around the orchard in order to analyze in more detail the spatial variability of wind speed within the orchard. The movable station was installed at 13 different sites (M1 to M13, Fig. 1) for an average period of one week each. It recorded half-hour averages of wind speed and direction, and these values were compared with the simultaneous records collected at the landmark station H1 using simple linear regression analysis (referred to as on-farm regressions), considering the H1 records as the independent variable. The Measure-Correlate-Predict technique (Derrick, 1993;Sánchez et al., 2011) was used to estimate long-term wind speed at the 13 measurement sites. The technique is based on relating the short-term wind speeds measured at the prediction site to those recorded at a long-term reference site. Different correlations were established at each measurement site for the predominant wind speed directions.
Additional simple linear regressions were performed between the half-hour records of wind speed at the SIAR reference station on one hand, and at the different measurement points within the orchard (H1, H2, and M1 to M13), on the other. These were referred to as off-farm regressions. In this case, the SIAR records were used as independent variable and the onfarm measurements as dependent variable.
The wind speed variability was used to estimate daily ETo, following the FAO Penman-Monteith approach (Allen et al., 1998) for the period 2004 to 2009 and at the different on-farm measurement sites (H1, H2 and M1 to M13). For these computations, daily meteorological variables recorded at the reference SIAR station were used at each measurement site, except for the wind speed, whose values were estimated using the local linear relationships.

Irrigation requirements
Crop evapotranspiration for standard (ETc Std ) and regulated deficit irrigation (ETc RDI ) conditions was estimated for the different crops of the study orchard following Eqs.
[2] and [3], respectively. Average ETo was determined for each cropping zone ( Fig. 1) accounting for the wind speed spatial variability and for the period of [2004][2005][2006][2007][2008][2009]. In this work the Kc values estimated by García-Vera & Martínez-Cob (2004) were used. These authors used the single Kc FAO-56 method (Allen et al., 1998) adapting them to the local climatic conditions of the Ebro river basin. Following García-Vera & Martínez-Cob (2004), the four crop stages required for the application of the FAO 56 methodology were defined as: 1) initial stage, from bud swelling to start of flowering; 2) development stage, from flowering to pit hardening; 3) mid-season stage, from pit hardening to ten days after harvest and 4) lateseason stage, from ten days after harvest to leaf fall.
In addition to the crop stages, fruit growth stages are commonly used to select the appropriate timing of RDI practices (Naor, 2006): 1) stage FI, from bloom to beginning of pit hardening; 2) stage FII, from beginning to end of pit hardening; 3) stage FIII, from pit hardening to fruit ripening (harvest); and 4) stage FIV, from harvest to leaf fall (postharvest), divided into early and late postharvest phases (before and after September 1). A seasonal RDI schedule results from the overlapping of crop and fruit growth stages, and from the use of crop coeff icients and def icit irrigation coefficients (Eq. [3]).
Local observations of the duration of the phenological stages for 21 cultivars of cherry, 6 cultivars of apricot, 12 cultivars of EMP, 11 cultivars of MMP and 7 cultivars of LMP were provided by the orchard irrigation manager for 2008 and 2009. The manager followed the BBCH methodology (Bleiholder et al., 1989) to establish the different crop and fruit development stages. Average durations were determined for the crops and for the two years of available data.
Reduction coefficients for the RDI irrigation strategies (Kr RDI ) were established following Girona et al. (2003Girona et al. ( , 2005 Table 1 presents the Kr RDI values summarized from the cited works. For cherry, apricot and early maturing peach, the RDI strategy reduced water application only at postharvest stage (FIV). For medium and late maturing peach, the adopted RDI strategy reduced water application at fruit growth stages FII and FIV (pit hardening and postharvest, respectively).
Net irrigation requirements for standard and RDI strategies (NIR Std and NIR RDI , respectively) were determined by subtracting the daily effective precipitation from daily ETc or ETc RDI . The daily effective precipitation was determined as 75% of daily precipitation (Dastane, 1978). Gross irrigation requirements for both standard (GIR Std ) and RDI (GIR RDI ) strategies were determined assuming an application efficiency of 90%, characteristic of drip irrigation systems (Clemmens & Dedrick, 1994). Gross irrigation requirements for both strategies were compared with observed irrigation water applications based on the criteria of the farm manager.

Scheduled and measured irrigation water application
The farm manager based daily irrigation scheduling on: 1) the data published by the SIAR network of agrometeorological stations; 2) the gradual implementation of RDI in order to limit pruning and to improve fruit quality; 3) an inspection of the crop status; 4) the readings of some soil water sensors installed at the orchard; and 5) fertigation requirements. According to the manager, fertilizer application occasionally required water in excess of crop water requirements, particularly at the beginning of the season. The daily time table of the irrigation schedule, scheduled irrigation time (SIT, hours) and the nominal water flow of each subunit valve were registered at the orchard manager's computer. Weekly scheduled irrigation depth (SID, mm) was obtained for each cropping zone in 2004, 2005 and 2006. Daily SID data was obtained for each valve in 2007, 2008 and 2009. Ten volumetric flow meters were installed at specific control points of the water delivery network, two per cropping zone, in 2008 and 2009. Control points were named C1-C10. Each flow meter was installed at the upstream end of a drip irrigation hose. The flow meters were equipped with a pulse emitter (one pulse was emitted each 10 L of flow) and a data logger. These data allowed estimating the temporal variability of irrigation discharge at each control point. Two types of coefficients of variation were determined: intra-irrigation (VC IA ) and inter-irrigations (VC IE ). Measured irrigation times (MIT) were compared with those scheduled by the orchard manager (SIT). The estimated irrigation depth was also compared with the computed standard and RDI gross irrigation requirements.

Irrigation system evaluation
The irrigated area containing each of the ten control points was selected to perform an irrigation system evaluation. Evaluations were performed following the ITRC rapid procedure (Burt, 2004). This methodology was designed to identify problems due to pressure differences between emitters and other causes of nonuniformity (clogging, manufacturing variation, ageing and friction). Measurements must be taken across an entire irrigated area (defined as a subunit in the irrigation evaluation methodology). The methodology is summarized in the following four steps: 1. The emitter discharge equation was determined for each subunit (Eq. [4]). The emitters installed at the orchard were turbulent (non pressure-compensating). As a consequence, they showed a strong pressure-discharge response. The individual flow rate (q) was measured at 16 contiguous emitters located at the upstream end of the hose. Measurements were performed at the operating pressure (p) and half of the operating pressure (Burt, 2004). The x exponent and the k parameter were determined by regression for the group of emitters: 2. The low-quarter distribution uniformity due to variations in emitter pressure (DU lqΔP ) was determined for each subunit (Eq. [5]). For this matter, pressure measurements were performed at four tree lines distributed along the subunit. At each of the two hoses irrigating the selected tree lines, pressure was measured at four locations.
Average of low -quarter estimated flows Average of all estimated flows 534 N. Zapata et al. / Span J Agric Res (2013) 11(2), 529-546 3. The low-quarter distribution uniformity due to other causes (DU lqOthers ) was determined following Eq.
[6]. Emitter flow rates were measured at three hoses within each subunit (Burt 2004). DU was determined at each location (DU i ), and DU lqOther was estimated as the average of the three locations using: where n is the number of emitters per tree. 4. The global low-quarter distribution uniformity (DU lqGlobal ) was determined following the statistical method proposed by Clemmens & Solomon (1997) (Eq. [7]): where K a is a factor that depends upon the type of data distribution. As suggested by Burt (2004), a typical value of 1.27 was used.

Stem water potential
Stem water potential (SWP) was measured at solar midday using a pressure chamber (Soilmoisture Equipment Corp., Santa Barbara, CA, USA), according to McCutchan & Shackel (1992). Samples were collected at the surroundings of the ten control points during the 2008 and 2009 irrigation seasons. At each control point, six mature leaves located on the shaded side of the central part of the third and fourth trees along the line (three leaves per tree) were measured between 12:00 and 13:00 UTC. Coverage of the 2008 season was partial, with measurements starting on July 10.

Soil physical properties
Soil stoniness thematic units can be described as follows: a) low stoniness (< 5%) for which average and standard deviation of RAW were 113 mm and 31 mm, respectively; b) medium stoniness (5%-20%) for which average and standard deviation of RAW were 70 mm and 7 mm, respectively; and c) high stoniness (> 20%) for which average and standard deviation of RAW were 41 mm and 15 mm, respectively. Due to the sampling density and the large spatial variability of RAW, it was not possible to map this variable with more detail. The magnitude of RAW spatial variability was similar to that reported in other areas of the central Ebro River basin (Playán et al., 2000;Lecina et al., 2005). The lowest observed values of RAW are compatible with the usual local irrigation requirements and practices (daily replacement of the crop water requirements), avoiding important deep percolation losses. The same can not be said about large precipitation events, which will result in strong differences in effective precipitation and deep percolation. Differences in RAW should be accounted when defining a RDI strategy since the level of deficit irrigation are also dependent on soil characteristics .
The comparison of the boundaries of soil units and the different cropping zones allow concluding that soil information was not used at the orchard design and at the irrigation system design phases. As a consequence, the farm manager's irrigation scheduling can not adequately accommodate the variability of soil water properties within the orchard (Boland et al., 2006;Fulton et al., 2011). Fig. 2 presents regression analyses of the four meteors used to determine ETo between the two landmark stations. Significant statistical differences could not be found in air temperature, air relative humidity and solar radiation. On the other hand, significant and important differences were found for wind speed between H1 and H2. The coefficient of determination of the linear regression, R 2 , was 0.93. However, wind speed at station H2 was (on the average) 52% of the wind speed at H1. When reference evapotranspiration was determined at both stations, differences in wind exposition at the two landmark points resulted in relevant differences in evapotranspiration: ETo at station H2 was only 80% of ETo at station H1. Since significant meteorological differences were only found for wind speed, the spatial variability of reference evapotranspiration within the orchard relied on the characterization of the spatial variability of wind speed.

Spatial variability of wind speed
The largest half-hour meteorological data series available in the area correspond to the SIAR reference station: from January 2004 to September 2009. Martínez-Cob et al. (2010 reported that the predominant wind directions in the Ebro river basin are West-Northwest and East-Southeast (locally named Cierzo Limitations to adopting regulated deficit irrigation in stone fruit orchards: a case study and Bochorno, respectively). Thus, the wind speed values recorded at the SIAR station were divided in two groups according to their respective wind direction (WD). Wind directions 180°-360°were classified as West (WD), while wind directions 0°-180°were classified as East (ED). A 64% of the data registered at the SIAR reference station showed a West wind direction, while 36% had an East wind direction. Considering all measurement locations, wind speeds from the West were larger and more intense than those from the East, except for the period between mid July and mid August 2008, when the East was the most frequent and intense wind direction. Table 2 presents the regression analysis between the landmark station H1 as independent variable, and the rest of wind speed measurement spots as dependent variables (on-farm regressions). Regression analysis for spots SIAR, M7, M8, M9, M11, M12 and M13 showed different models depending on the wind direction. Regression analyses for spots M2 and M8 indicated that local wind speed was not strongly related with wind speed at the reference sites SIAR and H1 (values of R 2 < 0.5). This finding may be attributed to the sheltering produced by the surrounding topography, resul-ting in local wind patterns. Similar results were obtained for spot M9 when the wind blew from the West, indicating that this point was equally sheltered from this wind direction.
On-farm regressions were used for prediction purposes. All regression statistics (not just R 2 , but also RMSE and MAE, data not presented) resulted better for H1 than for SIAR. The analyzed time series lasted from 2004 to 2009, and was only available for the reference SIAR measurement spot. Predicted values at the onfarm spots were obtained in two steps: 1) estimation of a complete wind speed time series for 2004 to 2009 at H1; and 2) estimation of a complete wind speed time series for 2004 to 2009 at the remaining on-farm spots.

Spatial variability of reference evapotranspiration
Wind speed variability was used to estimate reference evapotranspiration at the different wind station points and at the spots H1 and H2 for the period 2004 to 2009. Table 3  at the EMP zone were predicted from the daily average wind speed at points M1 and M11. The weight of each point was similar in all cropping zones. The only exception occurred at the cherry cropping zone. In this case, the M13 spot only accounted for 12% of the average wind speed and ETo, in agreement with the percentage of cherry trees surrounding this spot. The variability in wind speed and ETo between the measurement points was smoothed by the averaging process required for the estimation of average ETo in the cropping zones. If cropping zones would have been designed according to ETo, irrigation scheduling could have been much more responsive to the local water requirements. Adequate irrigation management requires the determination of local, standard crop water requirements. Our results have shown that the spatial variability of ETo needs to be taken into consideration at the study orchard in order to schedule irrigation at the cropping zones. In the current orchard design, the cropping zones include relevant ETo variability. For instance, average ETo values of 1,194 mm yr -1 (M6) and 1,468 mm yr -1 (M13) have been determined within the cherry area.

Irrigation requirements and scheduled irrigation application
The FAO-56 initial stage started on February 5 for all crops (Fig. 3). Table 4 presents the average length (days) of the four FAO-56 phases. The largest difference among crops for the duration of the initial stage was 10 days, observed between LMP and apricot. Large differences were observed for the duration of the development, mid-season and late-season stages, particularly between the extreme cases: LMP and Cherry (Table 4). Cherry presented the shortest development and mid-season stages, and the longest late-season stage. The duration of FII (from beginning to end of pit hardening) was determined as 29 and 63 days for MMP and LMP, respectively. The standard deviation of the duration of these phases was 7 and 5 days, respectively.
The irrigation strategies suffered relevant changes between 2005 and 2009. The adequate stages for deficit irrigation have been described for the crop species and cycles present at the orchard (Chalmers et al., 1981;Johnson et al., 1992;Torrecillas et al., 2000;Girona et al., 2005;Antunez-Barria, 2006). However, the deficit levels are cultivar, soil and fruit load dependent . Parallelism between SID and GIR RDI resulted evident during some periods of the analyzed seasons and crops, but in specific periods irrigation was not in deficit. The manager scheduled irrigation to avoid deficit irrigation at the stages sensitive to water stress (FIII). In general, during the beginning of the irrigation season, SID exceeded gross irrigation requirements. This was particularly evident during the 2007 season (Fig. 5), and can be due to an underestimation of early spring precipitation events or to large fertigation requirements.

RDI and water conservation
The manager's irrigation schedule resulted in water conservation respect to the standard strategy, ranging between 34 mm yr -1 for cherry and 347 mm yr -1 for apricot. However, introducing the RDI strategy would result in large water conservation respect to the observed water use, ranging between 260 and 351 mm yr -1 for EMP and MMP, respectively. The manager's irrigation schedule in 2009 used less water than the RDI strategy. However, the manager's schedule cannot be considered an RDI strategy.

Scheduled and measured irrigation time
A large variability could be observed in intraday scheduled irrigation time (SIT) for the different subunit valves of a given cropping zone. For instance, in 2008, the seasonal averages of the daily SIT CV were 20, 26, 45, 30 and 40% for cherry, apricot, EMP, MMP and LMP, respectively. This variability reflects the manager's reaction to the identified sources of variability in crop water availability and requirements.
In order to compare MIT and SIT for each crop, MIT was the average of the two measurement points, while SIT represented the average of the two irrigation subunit valves. During the 2008 irrigation season, SIT was in most of the cases higher than MIT. The largest diffe-rences (11.2%) were observed for MMP, and the lowest differences were observed for apricot (1.2%). In 2009, SIT was higher than MIT for cherry (data not showed), apricot (data not showed) and medium maturing peach (Fig. 6c). However, large negative differences were observed for EMP (Fig. 6b) and LMP (Fig. 6d), 38.5% and 25.6%, respectively. These differences seem related to operational problems during the opening and closing of the valves or to irrigation programming errors. Fig. 6 permits appreciating differences in detail for different crops and time periods. These data confirm the difficulty of managing irrigation in this complex orchard.

Discharge, pressure and uniformity of the irrigation system
The average discharge measured at each control point depends on the length of the irrigation hose and on the operating pressure. In order to analyse these data, hose discharge was standardized by the hose length. Since emitters are spaced 1 m, the standardised discharge equals the average emitter discharge. This value fluctuated between C10 and C3/C4 (2.6-5.1 L h -1 m -1 ) ( Table 6). The variability of discharge during the irrigation events largely depended on the control point. On the average, the CV of discharge within the irrigation events was 12%, while the CV of discharge between the different irrigation events was 10%. Space  and time variability in irrigation discharge constitutes a relevant source of variability in irrigation depth, seriously affecting irrigation performance. A detailed analysis of the water meter data revealed that the manager often adjusted the daily irrigation time of the different valves of the same cropping zone to correct the differences in discharge and its time variability. As reported in Table 6, the average irrigation discharge at C8 was significantly larger than at C9. In response to this, the manager increased irrigation duration and frequency in C9 respect to C8 (data not presented). The same pattern was observed at the two measured LMP cropping areas (C2 and C3, Table 6). On the other hand, the discharge differences between the two controlled apricot cropping areas were not effectively corrected. Dichio et al. (2007) concluded that the success of an RDI strategy heavily relies on the adequate design and management of drip irrigation systems. Our results indicate that, despite the manager's efforts, more control (flow meters, pressure regulators) would be required at the irrigation network in order to effectively regulate deficit irrigation in the orchard.
The manufacturing year of the emitters was 2001 for subunits C2 to C9, and 2004 for subunits C1 and C10. The average value of the x exponent was 0.46, reasonably close to the theoretical value of 0.50. Very relevant differences were detected on the emitter operation pressure. Irrigated areas C1, C4 and C5 showed very low working pressure (80, 70 and 57 kPa, respectively). Differences in pressure in the same cropping zone were very relevant. The results of the distribution uniformity analysis indicate that area C9 resulted in poor DU lq (0.73 for the tree and 0.72 for the emitter). Non-uniformity in this irrigated area was mostly due to differences in pressure (DU lqΔp = 0.73). Area C5 presented a moderate DU lq (0.85). The rest of irrigated areas showed high distribution uniformities, either referred to a single emitter or to all the emitters irrigating a tree. "Other" causes of non-uniformity were not relevant at the experimental orchard (tree DU lqOther ranged between 0.96 and 0.99). These values seem to indicate that the system is showing adequate ageing. However, Hanson et al. (1996), andBurt (2004) indicated that there is not a clear correlation between the age of the system and the DU lqOther .
The values of DU lq determined at the study orchard resulted higher than other references in the literature. In fact, the average tree DU lq at the orchard (0.92) was higher than the average DU lq of 0.73 reported by Hanson et al. (1996), or the average DU lq of 0.75 repor-ted at the Cachuma RCD (1994). Our experimental results were also higher than the average DU lq of 0.86 reported by Burt (2004) for drip irrigation systems. The differences in uniformity referred to the emitter or to the tree were in all cases very small. The large observed number of emitters per tree (5-9) was not required to attain adequate uniformity.

Assessing crop water stress
In general, stem water potential measurements at both control points of a given cropping zone showed similar patterns, with the exception of MMP in 2008, EMP in August 2008 and apricot in June 2009 (Fig. 7). The periods of proposed deficit for each cropping zone in the RDI strategy are presented in the Figure by horizontal arrows. For peach trees, Marsal et al. (2004Marsal et al. ( , 2005 reported that minimum values of around -1.5 MPa at the end of Stage II are associated with detrimental effects on fruit size at harvest. Similar results were presented by Dichio et al. (2007) in peach and by Marsal et al. (2009) in cherry. There are references in the literature pointing out that the threshold limits of SWP are variable between crops, crop development stages, soil characteristics and fruit loads. In this work, SWP = 1.5 MPa was used as a threshold limit of water stress (this threshold is represented in Fig. 7 by a horizontal dashed line). In cherry, crop water stress was important from early June to early July 2009, but no relevant stress was observed at the later postharvest period (August and September) of 2009. In general, the SWP measurements agree with the irrigation depth applied to the crops (Figs. 4 and 5). In apricot, crop water stress was observed in 2008 from mid July to mid August and in 2009 from mid June to mid July. SWP results for cherry and apricot showed some level of water stress during short periods of the recommended RDI periods. Although Fig. 5 displayed more deficit for EMP in 2009 than in 2008, SWP measurements showed less stress in 2009 than in 2008. These results could be explained by the abovementioned discrepancies between SIT and MIT. For MMP in 2009 water stress occurred at stage FII of fruit growth, and stress was maintained at the limit threshold during a large part of the rapid fruit growth (stage FIII). These results are compatible with the irrigation scheduling applied to MMP in 2008 and 2009 (Fig. 5). For LMP, slight crop water stress was found in 2009 only at a short period of stage FII of fruit growth (again in agree-ment with the schedule presented in Fig. 5); no crop water stress was identified after this period.
In general, the FII stage (fruit growth) was only partially exploited in MMP and LMP. Only at the end of this period water stress became evident in both cropping zones. Postharvest water stress appeared in cherry, apricot and EMP, particularly at the beginning of this phase. However, water stress was not present at the late postharvest period. Stress could be observed in water sensitive periods of EMP and during the whole cycle of MMP.

Conclusions
The case study has shown evolution towards deficit irrigation between 2004 and 2009. However, the complexities derived from the spatial variability of environmental factors (soils and meteorology) and irrigation performance limited adoption of a formal RDI strategy. The lack of consideration of environmental variability at the design phase and at the management practices derived in reactive water management. In the light of the results obtained at the case study orchard, the same can be said of irrigation performance. Design decisions, such as distributing varieties in the irrigated areas according to their phenology, using pressure compensating emitters, or installing pressure regulators and volumetric meters at the valves would have led to higher irrigation performance levels. The case study has illustrated that distribution uniformity is a requirement for high irrigation performance, but it does not guarantee it. In fact, the evaluated irrigation areas are uniformly irrigated, but the variability in operating pressure (among the studied irrigated areas and in time) makes it complicated to adequately schedule irrigation for standard or RDI conditions. The manager minimised the effect of environmental and irrigation variability by modifying the daily irrigation time for each subunit valve.
According to the literature, full implementation of RDI at the study orchard would conserve additional water, lead to optimum yield quality and minimize pruning requirements. Practical procedures to compute standard and RDI irrigation requirements have been summarized for all the crops from the scientific literature. The comparison of these requirements with SDI indicates that orchard irrigation practices were shortfall but did not correspond to an RDI strategy. Stem water potential confirmed some of the trends announced by the previous analyses, and revealed additional traits. The regular use of a plant water stress indicator seems to be a requirement to succeed in commercial RDI operation.