Comparison of irrigation scheduling strategies for achieving water use efficiency in highbush blueberry

Abstract

A glasshouse experiment was carried out to assess the feasibility of applying partial rootzone drying (PRD) to highbush blueberries (Vaccinium corymbosum). A subsequent field experiment was established to assess four irrigation strategies with the aim of improving water use efficiency in blueberry production. Applying PRD to plants during a glasshouse experiment reduced stomatal conductance without reducing plant water potential. Hindered by high rainfall, a physiological response to PRD was not repeated in field grown plants. However, irrigation scheduled using a K_c (crop coefficient) curve constructed from Food and Agriculture Organization (FAO) 56 guidelines and post-harvest regulated deficit irrigation (RDI) delivered annual water savings of 0.8 ML ha⁻¹ and 1.3 ML ha⁻¹, respectively, compared with a total 3.6 ML ha⁻¹ applied using a ‘rule-of-thumb’ approach commonly adopted by Australian blueberry growers. These savings were achieved without reducing berry yield or quality. This study is the first to report on the feasibility of applying FAO 56 guidelines, RDI and PRD as strategies to maximize water use efficiency in highbush blueberry production.

Keywords:

• Vaccinium corymbosum
• blueberry
• irrigation
• regulated deficit irrigation
• partial rootzone drying

Introduction

The majority of Australia's blueberries are produced within the higher rainfall districts of subtropical north-eastern New South Wales (NSW) with the remainder produced in temperate southern Australia. Rainfall patterns across subtropical and temperate Australia can vary considerably seasonally and from year to year. In subtropical and temperate eastern Australia, the greatest risk of water deficits occurs through the spring and summer months (September to February) with the highest risk of extended droughts occurring under the influence of El Niño events. Over the coming decades, the impacts of climate change are forecast to exacerbate climate variability and water security issues across eastern and southern Australia (Australian Department of Climate Change 2009). Consequently, as the blueberry industry continues to operate and expand in Australia, the need to better define and implement ‘best practice’ irrigation will be coupled with a need for strategies that enable growers to adapt to potentially increased frequency and duration of water shortages.

Current irrigation practices applied to southern highbush blueberry (Vaccinium corymbosum L.) cultivars grown in Australia are almost entirely based on local grower experience with some influence from a limited composite of studies undertaken in the US (Wilk et al. 2009). Baseline estimates for southern highbush blueberry water requirements are 25 mm during low-demand periods to 38 mm during high-demand periods applied weekly to the cropped area (Bell 1982). From these estimates many growers have adopted a standard c. 4 L water per mature plant per day ‘rule-of-thumb’ approach to irrigation. However, anecdotal observations by local professionals working with the blueberry industry indicate that inputs vary among growers, with a tendency towards over irrigation.

The Penman-Monteith reference evapotranspiration (ET_o) equation (Monteith and Unsworth 1990; Allen et al. 1998) provides an effective tool for managing irrigation inputs. To estimate plant water requirements, ET_o is calculated using climatic or pan evaporation data and crop evapotranspiration (ET_c) is estimated by multiplying ET_o by an appropriate crop coefficient (K_c). Greater adoption of this approach by Australian blueberry growers has the potential to significantly improve water use efficiency across the industry. However, one of the barriers to adoption is an uncertainty among growers of which K_c values to use. This uncertainty arises as there have been relatively few K_c curves published for southern highbush blueberries and these tend to vary for specific cultivars grown under different cultural conditions (Byers & Moore 1987; Storlie & Eck 1996; Haman et al. 1997a, b; Yang et al. 2005). In addition, no research has been undertaken to determine K_c curves for southern highbush blueberry cultivars grown in Australia. A generic mid-season K_c value for berry crops, presented in the Food and Agriculture Organization's Irrigation and Drainage Paper 56 (FAO 56) (Allen et al. 1998), potentially offers a satisfactory starting point for Australian blueberry growers, but its suitability requires evaluation.

To date, blueberry irrigation research has mostly focused on utilizing irrigation to optimize yield performance. However, under conditions of limited water availability, such optimums may not always be achievable and maximizing water use efficiency to maintain production may, at times, become the overriding priority. Deficit irrigation strategies have been successfully applied in a range of horticultural crops with no to little negative impact (Fereres & Soriano 2007). For some crops, such as grapes, increased yields have been achieved (de la Hera et al. 2007).

There are two main approaches to implementing deficit irrigation: first, regulated deficit irrigation (RDI) (FAO 2002; Kriedmann & Goodwin 2005); and second, partial rootzone drying (PRD) (FAO 2002; Kriedmann & Goodwin 2005). Deficit irrigation occurs when the water applied to a crop is below ET_c requirements. RDI is implemented by imposing prescribed limits on soil moisture depletion and limits to irrigation inputs at specific phenological stages of the crop cycle. PRD involves irrigation inputs that are below ET_c requirements but irrigation is alternated between plant roots separated into drying and wetted zones. When successfully applied, plants respond to PRD by reducing transpiration through partially closing their stomata without loss of plant turgor (Kriedmann & Goodwin 2005).

Although there are a handful of studies in which negative results have been reported after applying PRD (Gu et al. 2004; O'Connell & Goodwin 2004, 2007; Neuhaus et al. 2007), there are a greater number of studies that demonstrate successful application of the technique to grapes (Dry et al. 1996; Loveys et al. 1998; Stoll et al. 2000; Chalmers et al. 2004) and many other horticultural crops including apple (Leib et al. 2006), pear (Kang et al. 2002), mango (Spreer et al. 2009), olive (Fernandez et al. 2006), tomato (Tahi et al. 2007) and raspberry (Grant et al. 2004).

Despite successes in other crops, there are no reports of PRD having been trialled with highbush blueberries. We report on the outcomes of a pilot PRD pot trial carried out with the specific aim of observing whether the desired physiological response of PRD could be initiated in southern highbush blueberry plants. We also report on the outcomes of a field experiment in which a participating blueberry grower's ‘rule-of-thumb’ irrigation management strategy was compared with irrigation scheduled using ET_c estimated by applying a K_c curve constructed using FAO 56 guidelines (Allen et al. 1998). Both RDI and PRD were also evaluated in the field trial.

Materials and methods

PRD pilot pot trial

A PRD pot trial was conducted at the Centre for Tropical Horticulture, Alstonville, NSW, Australia. Prior to establishing the experiment, 12-month-old highbush blueberry cultivar ‘Star’ plants were transferred to and grown in 20 cm (6 L) pots. Each pot was divided into two chambers by an impermeable barrier and durable plastic bags were inserted into each chamber (plastic bags were punctured to permit drainage). The root ball of each plant was washed to remove adhering potting mix and split without damaging the crown. The two sides of the separated root ball were positioned on either side of the divider. Plants were grown in a glasshouse for a further 13 months (December 2007 to February 2008) in washed river sand mixed with 10 g complete slow release fertilizer pellets. During establishment, 5 g slow release fertilizer pellets were applied at 8-week intervals. Irrigation was controlled by automatic timers set to run for 5 min at midday and early evening each day (early evening only during winter months) with water delivered via two 1 L h⁻¹ drip emitters with one emitter installed on each side of the split pot.

Four treatments were applied throughout the experiment:

1.	full irrigation (FI=plants watered on both sides of the pot)
2.	alternating PRD (PRD=irrigation and drying alternated between each side of the pot)
3.	fixed PRD (FPRD=irrigation and drying each fixed to one side of the pot)
4.	stressed (plants watered on both sides of the pot but exposed to repeated stress and recovery cycles imposed by withholding water until plants began to wilt at which point irrigation was resumed).

Treatments were replicated four times in a randomized complete block and run over a 5-week period. During this period, the moisture content of the potted sand was measured daily using a single portable time-domain reflectometer (TDR) sensor with 12 cm probes (Campbell Scientific, Australia). Plants were irrigated as described above but with emitters installed on the irrigated sides of pots only. Mid-morning stomatal conductance and midday leaf water potential were measured for all plants at the start and completion of each stress and recovery cycle (as applied to the stressed treatment plants).

Field irrigation experiment

Trial site

A field experiment was established in a leading commercial blueberry orchard near the town of Wollongbar, north-eastern NSW, Australia (28.80^oS, 153.38^oE, 142 m ASL). The region is influenced by a subtropical climate with the nearby Alstonville weather station (11 km from study location) recording mean annual rainfall of 1817 mm (median 1683 mm) and mean annual pan evaporation of 1570 mm (1963–2010) (Australian Bureau of Meteorology 2010). The lowest rainfall typically occurs from July to January and the wettest months occur from February to April (Fig. 1). The formation of rainfall deficits is more common through winter, spring and early summer. Rainfall deficits tend to be more severe during El Niño years of which 11 events have occurred since 1963 (Australian Bureau of Meteorology 2010).

Figure 1  Rainfall and evaporation statistics recorded at Alstonville Tropical Fruit Research Station (28.85°S, 153.46°E, 140 m ASL) for the period between 1963 and 2010. Source: Australian Bureau of Meteorology (2010).

The trial site is characterized by basaltic clay-loam red Ferrosol soil (Isbelle 2002) of the Wollongbar series (Morand 1994). Soil bulk density measured from 36 soil cores taken from within soil mounds averaged 1.0 g cm⁻³ (between 10 cm and 40 cm depth) and 1.1 g cm⁻³ under soil mounds (between 40 cm and 100 cm depth). Applying the soil moisture characteristic model of Williams et al. (1983); soil group 7), the theoretical drained upper limit was estimated at 0.4 m³ m⁻³, plant available water content (AWC), i.e. water available between −0.1 MPa and 15 MPa, was estimated at 0.12 m³ m⁻³ and readily available water content (RAW), i.e. 50% of AWC, was estimated at 0.06 m³ m⁻³. During this study, core samples collected 48 h after saturating and draining the soil averaged 0.36 m³ m⁻³. Field capacity was therefore adjusted to 0.36 m³ m⁻³ and the theoretical AWC and RAW values were subtracted to give an estimated permanent wilting point (PWP) of 0.24 m³ m⁻³ and refill point of 0.30 m³ m⁻³. These estimates were within the range of previous measurements recorded for Ferrosol soils at nearby locations (Marsh & Rixon 1991; Batten et al. 1994; Bell et al. 2005).

Experimental treatments

Five-year-old highbush blueberry ‘Star’ cultivar plants were used for the study. Plants were established under the farmer's standard pruning, irrigation and fertilizer management practices and planted at 3 m row×0.8 m in-row spacings in mounded soil rows with dimensions of 1 m base width×0.4 m high. When the soil mounds were formed they were covered with permanent woven plastic weedmat with the weedmat folded under the base-edges of the soil mound thus semi-isolating the mound from the adjacent grassed inter-row.

During the field experiment, four irrigation treatments were applied:

1.	farmer practice
2.	ETc (100% replacement of crop water use)
3.	RDI (50% ETc applied to both sides of plants)
4.	PRD (50% ETc applied to alternating sides of plants).

Treatments were replicated four times in a randomized complete block with each block formed within a single row and five buffer rows (buffer rows contained different cultivars) separating each block. Each plot consisted of 10 plants with two or more buffer plants on the periphery of each plot. The experiment was run from November 2008 to end of harvest, November 2009. RDI and PRD treatments were applied over late spring-summer between early-November 2008 and end-February 2009. From March 2009 onwards, RDI and PRD treatment plots reverted to irrigation inputs based on 100% ET_c.

Irrigation water was delivered to plants via 2 L hr⁻¹ drip emitters spaced at 40 cm intervals (standard farmer practice). During preliminary observational studies, soil moisture measurements indicated that this emitter spacing maintained relatively even wetting of soil mounds. A number of excavations were also made to visually assess the effective rooting depth of plants. The highest root density was observed within the top 0.4 m of the soil with fewer roots observed to a maximum 0.6 m depth. Plant roots were also found to occupy space to the edges of the soil mound. Based on the area of soil occupied by each plant, the total irrigated cropped surface area was estimated at 0.8 m² plant⁻¹ (c. 3000 m² cropped area ha⁻¹).

The farmer's existing irrigation regime was applied to the farmer practice treatment and involved a target of applying 4 L water per plant at each irrigation event with 5 L water per plant applied once each week as fertigation. From post-harvest (November), through the summer to early-autumn (March) vegetative growth stage, irrigation was scheduled four times weekly (equivalent to c. 25 mm per week). During the cooler mid-autumn and winter months, irrigation was applied twice weekly (equivalent to c. 15 mm per week). From flowering in mid-August through to end of harvest, target irrigation volumes were increased to 5 L per plant at each irrigation event and irrigation was scheduled four times each week (equivalent to c. 30 mm per week). During the final 6 weeks of the crop cycle (fruit expansion and harvesting), irrigation frequency was increased to five times each week (equivalent to c. 40 mm per week). This regime was not rigorously applied, however. The irrigation manager continued to apply their experientially derived perceptions to guide irrigation inputs on any single day as influenced by rainfall and physical examination of soil moisture. During the trial, the on-farm irrigation manager recorded the date, time and duration of each irrigation event. These data were used to calculate irrigation inputs.

Irrigation in the ET_c, RDI and PRD plots was initiated at the same time as each irrigation event in the farmer practice plots and input volumes were controlled by varying irrigation runtimes using automatic timers activated by a pressure switch. The timers were adjusted weekly to accommodate the expected number of irrigation events and estimated plant water requirements for the proceeding week. An automated weather station (AWS) (Environmental Information Systems, Australia) installed central to the experimental block was used to monitor climatic variables, rainfall and calculate daily ET_o (AWS uses the standard FAO 56 Penman-Monteith equation to calculate ET_o).

Plant water requirements were estimated by applying a K_c curve constructed using FAO 56 guidelines (Allen et al. 1998). The crop cycle was defined in three main stages (Fig. 2): K_c ph=post-harvest through the vegetative growth stage (21 weeks; November to March); K_{c w}=semi-dormancy stage with partial leaf fall (16 weeks; May to August); K_{c h}=peak water demand period from start of ripening through to end of harvest (5 weeks; October to November). A generic berry mid-season K_c value (K_{c mid}) of 1.05 was used as the initial reference for each stage with adjustments made to account for crop and management factors. Equation 1 was used to construct the adjusted K_c (K_{c adj}) curve.

1

Figure 2  K_c curve applied during the field experiment.

where:mulch adjustment coefficient, M_c=K_{c mid}×0.2 post-harvest adjustment coefficient, ground cover coefficient, ; where f_c=ground cover fraction

The plastic weedmat was assumed to reduce K_c by 20% (Allen et al. 1998) and was applied throughout each stage of the crop cycle. Observations have been made of plant water requirements in highbush blueberries rapidly declining by c. 30% post-harvest (Bryla & Strik 2007) and local growers also reduce their post-harvest irrigation inputs by around 30%. Consequently, during the post-harvest vegetative growth stage (K_c  ph) a 30% post-harvest adjustment (P_c) was applied from November to March. Under local conditions, during the cooler months, ‘Star’ cultivar is semi-dormant with partial leaf fall occurring. To compensate for this, adjustments for changes in the ground cover fraction (f_c) were made during the 16-week semi-dormant period (K_c w), when the ground cover fraction declined from an estimated 90% to 30% due to leaf senescence and partial leaf fall.

Adjustments were also made for rainfall infiltrating into the soil under the plastic weedmat. To simplify water budget calculations, 20% infiltration of the gross incident rainfall was assumed. This assumption was based on course estimates for rainfall entering the soil via stem flow (the portion of incident rainfall that infiltrates the soil via the trunk) (Levia & Frost 2003), holes cut into the plastic weedmat to accommodate plants and seepage through pores occurring at strand intersections within the woven plastic weedmat.

Soil moisture measurements

A Micro-gopher^© mobile capacitance probe (Odyssey, New Zealand) was used to manually record soil moisture. Prior to using the Micro-gopher^©, to record soil moisture values during the field trial, an insitu calibration was carried out. This was undertaken as recommended by Geesing et al. (2004) with comparisons made between soil moisture readings recorded with the Micro-gopher^© and 36 soil core samples taken adjacent to probe reading positions at depths between 10 cm and 100 cm across the study area. A regression analysis of Micro-gopher^© generated soil moisture values against the same obtained from soil cores returned an r²=0.93 (P < 0.01). The margin of error across 36 observations averaged ±0.1 mm with a maximum error of ±2.5 mm.

During the field trial, moisture was measured on a weekly cycle with most measurements recorded on the same day near to the same time once each week. Measurements were recorded at 10 cm intervals to a depth of 50 cm in each plot with both left and right sides measured in PRD plots. Soil moisture data are presented as means of measurements recorded across the four replicate plots for each treatment.

Midday leaf water potential

Midday leaf water potential was measured using a plant water status console (Model 3005, Soil Moisture Equipment Corp, US). Sunlit mature leaves were selected for measurements. Stem tips with two to three leaves attached were covered in a plastic bag (to slow transpiration) and excised from the plant. The excised stem with leaves attached was then placed immediately inside the pressure chamber. Compressed N₂ gas was released into the chamber until water appeared at the cut stem surface, at which point the inlet valve was closed and the pressure inside the chamber recorded. During the pot trial, duplicate values were recorded for each plant with stems excised from opposing canes. For the field experiment, single values were recorded for each plot.

Stomatal conductance

Stomatal conductance was measured using a leaf porometer (Model SC-1, Decagon, US). Fully opened sunlit leaves were selected and measurements were carried out on abaxial surface of leaves between 9:00 am and 10:30 am. Intervals between measurement events were dictated by the absence of rainfall and cloud cover.

Berry yield

Blueberries were manually harvested by professional fruit pickers. Three repeat harvests of ripe berries were carried out over a 4-week period commencing 6 October 2009. Berries were separated into premium and second grade classes and the weights of both recorded for each plant. The weights of both were combined to give a total yield and the weight of the second grade berries were expressed as a percentage of the total yield. The individual weights of a subsample of 50 berries from each plot at peak harvest (second harvest) were also recorded and averaged for each plot.

Statistical analysis

Where appropriate, data were fitted to a general linear model and means were separated by Tukey's honestly significant difference (HSD) (P=0.05). All analyses were carried out in the R environment (R Development Core Team 2009).

Results

PRD pilot pot trial

Drying of the potted sand was easily achieved within a few days when irrigation water was withheld from pots (Fig. 3A–D). This facilitated frequent drying cycles on alternating sides of PRD plants and permitted stress and recovery cycles to be completed on a weekly basis in stressed treatment plants. Depletion of water in non-irrigated soil was negligible once sand dried to 5% (Fig. 3C).

Figure 3  Volumetric soil moisture measured daily over the duration of the blueberry PRD pot experiment. Data points represent means (n=4). A, Fully irrigated. B, Partial root zone drying. C, Fixed partial rootzone drying. D, Plants exposed to repeated stress and recovery cycles.

At the commencement of the experiment, stomatal conductance was similar in all plants (Fig. 4). Within 4 days, stomatal conductance fell to an average 18.4 mmol m⁻² s⁻¹ in stressed treatment plants. This pattern was repeated at the completion of each stress cycle. At 264 to 300 mmol m⁻² s⁻¹, stomatal conductance in the PRD and FPRD plants (respectively) was well below the fully irrigated plants which scored a mean value of 370 mmol m⁻² s⁻¹. During the second drying cycle, the gap in stomatal conductance, measured between fully irrigated plants and plants exposed to PRD and FPRD, widened significantly (P<0.05) to around half that observed for fully irrigated plants. Stomatal conductance partially recovered in the stressed plants within 3 days of resuming irrigation with stomatal conductance rising to match values observed for PRD and FPRD plants. Similar patterns were observed during subsequent drying cycles with the only deviation being for FPRD plants whereby at the completion of the third cycle, stomatal conductance for FPRD plants was higher than for PRD plants but similar to fully irrigated plants. This pattern was repeated at the commencement of the fourth cycle but was significantly different (P<0.05) between all treatments at the completion of the experiment. A significant (P<0.05) decline in leaf water potential was observed in stressed treatment plants during the deficit stage of each stress and recovering cycle (Fig. 5) but differences did not occur between the other three treatments.

Figure 4  Abaxial stomatal conductance (g _s) measured in potted blueberry plants exposed to fixed irrigation (FI), partial root zone drying (PRD) and fixed partial root zone drying (FPRD) and stressed treatments. Data points represent means (n=4) and error bars indicate standard errors of the mean.

Figure 5  Midday leaf water potential (Ψl^md) measured in potted blueberry plants exposed to fixed irrigation (FI), partial root zone drying (PRD) and fixed partial root zone drying (FPRD) and stressed treatments. Data points represent means (n=4). Error bars are not shown as the standard errors of the mean were each<0.05.

Field experiment

Over the duration of the field experiment, the region experienced frequent and above average rainfall (annual total 2235 mm), with most of this rain falling between November 2008 and July 2009. While rainfall exceeded ET_o and estimated ET_c, after accounting for 20% rainfall infiltration into soil beneath the plastic weedmat, both cumulative ET_o and ET_c exceeded water entering the soil as rainfall (Fig. 6). There were also several periods through late spring to early autumn during which ET_o and ET_c exceeded rainfall for that period. An extended period with very little rainfall also occurred from mid-winter through to mid-spring.

Figure 6  Cumulative rainfall, estimated rainfall infiltration, ET_o and ET_c over the duration of the blueberry field trial.

Irrigation inputs for the 12-month season totalled 3.60 ML ha⁻¹ (equivalent to 1174 mm applied to the cropped area) under the farmer's regime and 2.80 ML ha⁻¹ (913 mm cropped area equivalent) for the ET_c treatment and 2.42 ML ha⁻¹ (790 mm cropped area equivalent) for both RDI and PRD treatments (Table 1). During the peak ET_o period (November to March), estimated plant water requirements totalled 536 mm of which an estimated 143 mm was met by rainfall infiltrating beneath the plastic weedmat. During this period, the farmer's irrigation inputs totalled 405 mm, inputs to ET_c plots totalled 355 mm, and RDI and PRD plots totalled 232 mm. This contrasts with the lowest evapotranspiration period (April to August) during which estimated plant water requirements totalled 172 mm, rainfall infiltration was estimated at 237 mm and the farmer's inputs totalled 263 mm with ET_c, RDI and PRD plots receiving 180 mm, mostly from mandatory once weekly fertigation events. Between flowering and final harvest (mid-August to early November), the farmer applied 506 mm in addition to estimated rainfall infiltration of 44 mm. Estimated plant water requirement for the same period was 347 mm with ET_c, RDI and PRD plots each receiving 378 mm.

Table 1  Irrigation volumes for each treatment during the field trial.

Stage of crop cycle	Treatment	Total (mm)	Total (ML ha^−1)	Daily average (mm)
Post-harvest ▸ vegetative growth November—March	Farmer	405	1.23	2.7
147 days; Kc 0.6; ETc 536 mm; 20% rainfall infiltration	ETc	355	1.08	2.4
143 mm	RDI/PRD	232	0.70	1.6
Bud develop ▸ semi-dormanncy April—August	Farmer	263	0.83	1.9
142 days; April Kc 0.5, May to August Kc 0.4; ETc	ETc	180	0.57	1.3
172 mm; 20% rainfall infiltration 237 mm	RDI/PRD	180	0.57	1.3
Flowering ▸ end harvest mid-August—November	Farmer	506	1.54	5.4
77 days; mid-August Kc 0.5 to Kc 0.8 in final four weeks;	ETc	378	1.15	4.1
ETc 347 mm; 20% rainfall infiltration 44 mm	RDI/PRD	378	1.15	4.1
Totals	Farmer	1174	3.60
	ETc	913	2.80
	RDI/PRD	790	2.42

Throughout the duration of the experiment, soil moisture in the farmer's plots remained relatively constant within a range at or near field capacity (Fig. 7). A similar pattern was observed for ET_c plots. Soil moisture in both these treatment plots demonstrated a slow declining trend from November through to late January. The decline in soil moisture was much more rapid in the RDI and PRD plots during this same period. Increased rainfall from early February and a resumption of irrigation inputs based on ET_c estimates from March resulted in a steady increase in soil moisture through the autumn and winter months. Between November and March, alternating wetting and drying patterns in the PRD plots were observed. However, frequent rainfall interfered with the clarity of this pattern.

Figure 7  Soil moisture measured at weekly intervals during the field experiment. A, farmer practice. B, ET_c treatment. C, RDI treatment. D, PRD treatment. Values represent mean (n=4) soil moisture summed to 40 cm depth (sum of soil moisture measured at 10, 20, 30 and 40 cm depth). FC = field capacity; Refill = soil moisture depletion threshold (0.3 m³ m⁻³) at which point the soil profile should be refilled.

At no point during the PRD and RDI application phase were significant differences in stomatal conductance and leaf water potential observed between treatments (data not shown).

Mean berry yield ranged between 1.59 kg plant⁻¹ in the PRD plots to 2.02 kg plant⁻¹ in the RDI plots but differences between treatment means were not statistically significant (Table 2). The percentage of the total berry yield graded as seconds ranged between 17.6% for the farmer practice plots and 19.4% for the PRD plots. Average berry weights ranged between 2.7 g for the farmer practice plots and 2.3 g for ET_c plots. Differences between means for both these quality parameters were not significant. Under the farmer's regime, WUE was 2.1 kg kL⁻¹ and a significant (P < 0.05) improvement in WUE was achieved under the RDI treatment (3.1 kg kL⁻¹). The ET_c and PRD treatments delivered a mean WUE value of 2.5 kg kL⁻¹.

Table 2  Blueberry yield, quality and water use efficiency (WUE). No significant differences were observed for yield, percent seconds or berry weight (n=4).

Treatment	Yield (kg/plant)	Seconds^a (% of total)	Berry weight^b (g)	WUE^c (kg kL^−1)
Farmer	1.97	17.6	2.7	2.1
ETc	1.84	15.1	2.3	2.5
RDI	2.02	14.7	2.5	3.1
PRD	1.59	19.4	2.4	2.5
Tukey HSD (P<0.05)	0.52	5.80	0.4	0.8
^aPercentage of berries graded as seconds.
^bMean weight of 50 berries per plot sampled at peak harvest.
^cWater use efficiency calculated as plant yield/total volume of water applied to each plant (10 plants per plot). WUE values shown are treatment means (n=4).

Discussion

During our field experiment, FAO 56 guidelines (Allen et al. 1998) for estimating plant water requirements (ET_c) were evaluated against the farmer's ‘rule-of-thumb’ approach. The field experiment commenced in November 2008 with a K_{c ph} value of 0.6 through to March 2009. During this period, the farmer's inputs were slightly above ET_c until early January when inputs were then slightly below ET_c through to the end of March. By necessity, irrigation water for each plot was drawn from a single water main. Consequently, the maximum water input for all treatments was limited by the frequency and duration of the farmer-initiated irrigation events. This resulted in a minor deficit of 38 mm developing in the ET_c treatment plots and was accompanied by an equivalent decline in soil moisture. Regardless, irrigation inputs in both treatments were close to estimated ET_c, the soil moisture data indicated that water inputs were adequate to maintain RAW and leaf water potential and stomatal conductance measurements did not indicate that plants were water stressed.

As autumn temperatures cooled, a K_c value of 0.5 was applied in the month of April. With further cooling and changes in leaf colour progressing to partial leaf fall (signalling the onset of semi-dormancy), the K_c value was reduced to 0.4 through May to mid-August. Estimated rainfall infiltration during this period exceeded ET_c by 65 mm and the farmer applied 263 mm irrigation water. Irrigation inputs totalling 180 mm in the ET_c, RDI and PRD plots were also well in excess of estimated plant water requirements. Most of this excess water was added via mandatory fertigation events which, as a standard practice, were initiated once each week throughout the trial. It was necessary to permit fertigation in all plots to ensure that each experimental treatment received nutrient inputs equivalent to the farmer practice plots. Had these fertigation events not occurred, total irrigation inputs could have been reduced by 0.33 ML ha⁻¹. This indicates that there is scope for growers to reconsider the need for frequent fertigation events through the winter.

From the commencement of bloom in mid-August, K_c values were increased in 0.1 increments at fortnightly intervals until reaching the ripening stage when an adjusted K_c value of 0.8 was applied through to the end of harvest. Estimated plant water requirements for this period totalled 347 mm with 378 mm water applied to experimental treatment plots and 506 mm applied to plots irrigated under the farmer's regime. Rainfall was relatively low through most of this period with only 6 mm estimated rainfall infiltration up to the final 2 weeks of the harvest period when 186 mm rain fell, taking the estimated total rainfall infiltration to 44 mm. Low rainfall during most of the period was probably one factor that contributed to the farmer's perception of irrigation requirements. Another contributing factor is that preventing fluctuations in soil moisture during ripening is considered critical to avoid fruit splitting. As such, the standard practice is to increase the frequency and volume of irrigation to maintain consistent soil moisture during fruit expansion and ripening (Strik et al. 2003). Without making adjustments for changes in actual plant water use, however, there is a risk of over-watering as was the case in this instance.

A second objective for the field trial was to evaluate the feasibility of applying RDI as a strategy to conserve water in blueberry production under a water shortage scenario. When applied to ‘Star’ cultivar between the November to February summer vegetative growth stage, RDI reduced total irrigation inputs by 1.18 ML ha⁻¹ and 0.38 ML ha⁻¹ compared with the farmer and ET_c treatments, respectively. With water inputs set at 50% ET_c, soil moisture progressively declined between November and January, finally falling outside the RAW range in the final week of January. A plant physiological response to the water deficit was expected by this stage but was not observed. Bryla & Strik (2006) found that, independent of cultivar, midday stomatal conductance in highbush blueberry plants decreased as midday plant water potential approached −0.6 to −0.8 MPa. Davies & Johnson (1982) observed that the critical midday water potential for stomatal closure in the more drought-tolerant rabbiteye blueberry (V. ashei) was −2.2 MPa. In the present study, midday leaf water potential never fell below −0.66 MPa and mid-morning stomatal conductance consistently moderated between the lowest measurement of 260 mmol m⁻² s⁻¹ and the highest of 396 mmol m⁻² s⁻¹ and no significant differences were observed between treatments for leaf water potential or stomatal conductance. The apparent absence of a physiological response to the water deficit may be explained by the possibility that actual RAW was greater than that estimated using the model of Williams et al. (1983). It is also possible that plants were accessing water deeper than the estimated 40 cm effective rooting depth. Even so, during this study an irrigation deficit was applied within prescribed limits without any apparent water stress and without negatively impacting on berry yield or quality.

Prior to this study, the concept of applying RDI to highbush blueberries may have been overlooked due to several studies having demonstrated the sensitivity of highbush blueberries to soil water deficits. For example, Haman et al. (1997a, b) reported that growth and yield of the highbush cultivar ‘Sharpblue’ were significantly reduced when the Ψ_soil (soil water potential) was maintained below −10 kPa. Matric potentials in soils that are at field capacity range between −10 kPa for coarse textured soils and −33 kPa Ψ_soil for fine textured soils (Cassel & Nielson 1986). As such, at first glance, the results of Haman et al. (1997a, b) seem to indicate that the performance of highbush blueberries declines when soil moisture is permitted to fall, even marginally, below field capacity. It should be noted, however, that soil moisture in the Haman et al. (1997a, b) study was maintained at either −10 kPa, −15 kPa or −20 kPa throughout the duration of the experiment, which ran for 3 years. This contrasts with the current study in which we propose that RDI could be applied as a short-term strategy when water resources become limited. In eastern Australia, the highest risk of water shortages occurs from late spring through summer, which coincides with the post-harvest vegetative growth stage of early and mid-season southern highbush blueberry cultivars such as ‘Star’. Bloom and fruit expansion are the phenological stages most sensitive to water stress (Ameglio et al. 1999; Mingeau et al. 2001). As such, RDI may have greater potential to cause negative impacts on yield if imposed at these stages. However, further research is required to determine whether varying prescribed levels of deficit within and outside the vegetative growth stage has application to maximise water use efficiency.

Our final objective was to evaluate PRD as a water conservation strategy for highbush blueberry production. The primary criterion for declaring a successful plant response to PRD is an observation of stomatal closure without loss of turgidity (Kriedmann & Goodwin 2005). This criterion defined the objective for our pilot glasshouse pot trial during which such a response was observed, as indicated by a significant reduction in stomatal conductance without a corresponding fall in leaf water potential. This result supports previous observations of isohydric behaviour in highbush blueberries made by Ameglio et al. (1999). For such a response to have occurred, the plants used in the trial needed to have at least some capacity to translocate water from the irrigated roots to other parts of the plant. This insight is important in that it challenges the standing hypothesis which states that highbush blueberries do not have the capacity to translocate water laterally (Abbott & Gough 1986; Strik et al. 2003). This hypothesis originates from studies by Gough (1984) and Abbott & Gough (1986). The critical difference between our split-pot experiment and theirs was that we did not damage the crown when separating the roots. The current understanding of lateral hydraulic translocation in woody plants is that it mostly occurs overnight with water from roots in wet soil moving in an axial and circumferential flow through the lower stem, to opposing roots with lower water potentials. This translocated water is then lifted to the canopy when transpiration resumes (Smart et al. 2004; Burgess & Bleby 2006). By splitting stems 5 cm vertically through the basal crown, Gough (1984) and Abbott & Gough (1986) may have removed the plant's ability to translocate water and nutrients from roots on the irrigated side to roots on the non-irrigated side.

Gough (1984) and Abbott & Gough (1986) also fixed irrigation to one side of the plant and, after 6 months, observed that roots on the dry side were severely damaged. Others have also observed that highbush bluberry plants perform poorly with incidental fixed partial watering of roots (Shelton & Freeman 1989; Strik et al. 2003). These are reasonable observations considering that extended exposure of roots to dry soil would likely cause vascular damage which in turn would remove the roots’ ability to transport water and nutrients. Unlike fixed partial wetting, however, PRD involves alternations of wetting and drying to each side of the root system with alternations initiated before the soil has dried to the point where vascular damage might occur. Root damage by desiccation would also remove the ability of roots to produce and transport abscisic acid (ABA), a plant hormone that acts as the primary mechanism driving the plant's physiological response to PRD (Stoll et al. 2000; Davies et al. 2002; de Souza et al. 2005; Dodd et al. 2006; Liu et al. 2006). With each PRD cycle, ABA is produced in roots on the drying side of the plant and when irrigation is resumed to these roots, the water that they transport into the plant carries a pulse of ABA to the leaves. Upon arrival in the leaves, the ABA signals the stomata to reduce their aperture. During our pot trial, FPRD plants exhibited a physiological response similar to the PRD plants. This response indicates that, initially, the roots probably continued to receive water translocated overnight from the irrigated roots, produce ABA and transport both into the above-ground parts of the plant when transpiration resumed at daylight. During the final 2 weeks of the experiment, however, the stomatal response in FPRD plants diminished which may have been an early symptom indicating the onset of root damage due to desiccation.

While results from the pilot glasshouse experiment demonstrated a stomatal response to PRD in southern highbush blueberry plants, we were unable to stimulate the same response to PRD in field grown plants. This result may be attributed to unfavourable climatic and soil conditions. Rainfall throughout the duration of the PRD application period (November to February) was relatively frequent. While most of this rainfall would have been deflected by the plastic woven weedmat, some also infiltrated the soil as indicated by rises in soil moisture coinciding with rainfall events. In between frequent rainfall events there was also persistent cloud cover which reduced ET_o and the rate of soil drying. These conditions made it difficult to maintain continuity in soil drying cycles in the PRD treatment plots. No rain fell during a 3-week period in January at which time ET_o also accelerated. Soil moisture data from this period indicated that only in the final week of a 3-week cycle did soil moisture begin to demonstrate a pattern consistent for what would be expected with alternating irrigations under PRD. This period was the most likely time during the trial in which to observe a plant response to PRD. Soil moisture on the drying side fell within a band at which a stomatal response to PRD was expected but no such response was observed.

Clay-loam Ferrosol soils characteristically have good drainage but they also have a high water-holding capacity comparable with heavier clays (McKenzie et al. 1999). As such, soil water is available to plants for a longer duration than in, for example, sandy soils (McKenzie et al. 1999). Kriedmann & Goodwin (2005) provide several examples where small or nil responses to PRD occurred when PRD was applied to plants grown in clay soils. They speculate that the most likely explanation for these poor responses is due to soil moisture depletion on the drying side of plants being too sluggish to generate distinctive and repeated pulses of ABA. This provides a plausible explanation for both the successful application of PRD in the glasshouse experiment, where wetting and drying cycles were alternated every 3–4 days, and the lack of stomatal response to PRD in the field experiment, where soil drying may have been too slow to permit alternating wetting and drying cycles of sufficient frequency. The conflicting results between the pot and field experiment highlight an opportunity for re-evaluating PRD with highbush blueberries grown in soil with low water holding capacity (e.g. a sandy soil) and under more favourable or controlled climatic conditions than could be achieved during our field study.

Conclusion

The results from this study indicate that using FAO 56 guidelines to construct a K_c curve, commencing with a generic berry K_c mid of 1.05, can be recommended for adoption as a strategy to improve water use efficiency among Australian blueberry growers. We summarized these guidelines into an equation to save growers and extentionsists the trouble of having to read and interpret FAO 56. Using this equation to adjust for crop and site conditions specific to our field trial, we arrived at an adjusted K_c curve for the southern highbush blueberry cultivar ‘Star’ of 0.4 K_{c w}, 0.8 K_{c h}, 0.6 K_{c ph}. The suitability of this K_c curve was validated by observations recorded during the field trial. We stress, however, that rather than applying these exact K_c values, growers should use the equation defined in this study to construct a K_c curve adjusted to suit their situation.

We found that under the farmer's ‘rule-of-thumb’ regime plant water requirements were over-estimated during the cooler autumn and winter months and during spring flowering through to harvest. During winter, most water inputs were via weekly fertigation and so it is clear that this practice needs to be re-evaluated. Reducing irrigation in winter would also leave more water in storage in preparation for unforeseen water shortages, the highest risk of which occurs during the east-Australian spring and summer. Should a blueberry farmer be faced with water shortages under such conditions, our results indicate that RDI could be applied to southern highbush blueberries during the post-harvest vegetative growth stage. However, a grower should apply RDI within the prescribed limits of replacing water used by the crop at a rate of 50% ET_c with additional water applied when required to prevent soil moisture falling to a critical stress point (i.e. past full depletion of the RAW).

Acknowledgements

This work was funded by the Australia Centre for International Agricultural Research (ACIAR), Southern Cross University and Industry and Investment NSW under ACIAR project SMCN2003/035. Lisa MacFadyen is acknowledged for contributions as an adviser, Stephen Morris is acknowledged for biometric services and Samuel North and Brian Dunn are acknowledged for reviewing the manuscript.