The olive tree (Olea europaea L.) is a characteristic species of the Mediterranean basin, which has traditionally been managed under dryland conditions. However, many studies have shown the benefits of irrigation on yield (Patumi et al., 2002; Moriana et al., 2003; Tognetti et al., 2007; Martín-Vertedor et al., 2011). For this reason, most of the commercial olive orchards in South America nowadays have been established at fairly high densities with drip-irrigation systems (Correa-Tedesco et al., 2010). Hedgerow orchards at super-high densities are also becoming a more common training system (Connor et al., 2014).

Despite yield gains at the farm level, increasing water scarcity in many regions has led to increased competition for water with non-agricultural users (Fereres et al., 2003). If less water is available, farmers should look toward increasing water productivity (production per unit of total water applied) through the optimization of irrigation management (Fereres and Evans, 2006; Iniesta et al., 2009; Fereres et al., 2014). For olive orchards, the regulated deficit irrigation (RDI) (Tognetti et al., 2005, 2007; Iniesta et al., 2009; Gómez del Campo and García, 2013) is the most commonly used irrigation strategy and consists of imposing water stress during phenological phases that are relatively insensitive to water deficit. Goldhamer (1999) reported that the pit hardening phase is the least sensitive to water deficit, and recommended the adoption of RDI, restricting irrigation during this phase. RDI strategies have achieved savings of around 20% of total water applied without reducing fruit yield (Goldhamer, 1999; Gómez-del-Campo, 2013). Additionally, these studies indicated that the oil content was not affected by the decrease in total amount of water applied. Moreover, Iniesta et al. (2009) observed that the water productivity for oil production was tripled when there was a 25% decrease in total applied water. Similarly, Correa-Tedesco et al. (2010) indicated that the greatest water productivity (21.3 kg mm^-1 ha^-1) was observed when applying water between 51 and 52% of actual evapotranspiration (ETc).

Traditionally, ETc is computed using grass reference evapotranspiration (ETo) multiplied by grass-reference-based crop-specific coefficients (Kc). ETo is estimated using the Penman–Monteith combination equation, but there is uncertainty on how to select the appropriate values of Kc. In this case, Kc values are empirical and often not adapted to local conditions (Ortega-Farias et al., 2009; Poblete-Echeverría and Ortega-Farías, 2013). The Kc in olive orchards depends on aspects of canopy architecture such as orientation (Connor et al., 2014), ground cover (Martínez-Cob and Faci, 2010), and the interactions of climatic conditions, soil type, cultivars and irrigation management practices (Ortega-Farías and López-Olivari, 2012). Due to this potential difficulty, recent research in olive trees has suggested using stem water potential (Ψ_stem) to monitor plant water status and for scheduling water application (Moriana et al., 2012). Despite that some studies indicate that plant water status measurements could be strongly affected by the environment, which would question their usefulness as an irrigation scheduling tool (Corell et al., 2016), the water potential is a measurement commonly used as a reference in the description of water stress level (Moriana et al., 2012).

Irrigation cut-off strategies using water potential thresholds have been suggested for several researchers in prune, vineyards, and olive orchards (Lampinen et al., 2001; Girona et al., 2006; Moriana et al., 2012; Trentacoste et al., 2015). These strategies consist of suppressing irrigation completely during a given phenological phase which is insensitive to water deficit, and reestablishing irrigation only when a threshold value of Ψ_stem is reached. These strategies are easy to use for most farmers, since little knowledge is necessary about olive physiology in response to water stress.

In the literature, there is little information regarding irrigation strategies using Ψ_stem in olive trees. Moriana et al. (2012) observed that fruit yield decreased 30% in olive trees (cv. Cornicabra) that were irrigated when Ψ_stem fell below -2.0 MPa versus trees that were irrigated based on a -1.2 MPa threshold. Also, Fernandes-Silva et al. (2010) observed that olive trees (cv. Cobrancosa) irrigated when Ψ_stem reached -6.0 MPa had reductions greater than 50% in comparison with that of trees maintained under a Ψ_stem of around -3.0 MPa throughout the season. Ghrab et al. (2013) observed that the dry olive weight decreased significantly with Ψ_stem around -3.0 MPa. However, fruit yield for olive trees (cv. Frantoio) irrigated when the Ψ_stem dropped below -2.5 MPa, was statistically similar to the control (Ψ_stem threshold between -1.2 and -1.5 MPa) (Trentacoste et al., 2015). Correa-Tedesco et al. (2010) indicated that water deficit (Ψ_stem = -2.5 MPa) did not affect fruit weight. According to Dell’Amico et al. (2012), the lower yields can be attributed to the effect of water stress on fruit size. Finally, the effect of water deficit on yields depends on crop load, and is much more sensitive in years of high olive fruit load (Martín-Vertedor et al., 2011). This generates uncertainty regarding the use of Ψ_stem thresholds for irrigation in super-high density olive orchards (Naor et al., 2013). Due to these uncertainties, the objective of this study was to evaluate the yield and water productivity responses to irrigation cut-off strategies applied after fruit set using Ψ_stem thresholds in a super-high density olive orchard (cv. Arbequina).

Our region has a cool Mediterranean-type climate with about 620 mm of rainfall occurring during the winter months. During this study, weather behaved according to the expected conditions in the area, with maximum atmospheric demand during December and January. Due to low rainfall in summer, we found that irrigation was necessary to maintain fruit yield and total oil yield, and that irrigation cut-offs based on Ψ_stem thresholds was a practical and user-friendly method of scheduling irrigation in super-high density orchards. However, we recognize that optimal midday stem water potential may vary somewhat by region due to climate variables such as vapor pressure deficit and temperature (Corell et al., 2016), or due to soil type.

Irrigation of the T₁ treatment averaged 230 mm over the four growing seasons. For these seasons, Ψ_stem mostly ranged between -1.4 and -2.0 MPa, indicating that trees were in a null to mild water stress condition. This can be established because despite that thresholds from -1.0 to -1.5 MPa have been suggested as adequate to satisfy olive tree water requirements (Dell’Amico et al., 2012), values lower than -2.0 MPa can occur even in well-watered trees during the summer under high vapor pressure deficits and high crop load conditions (Ben-Gal et al., 2010; Ortega-Farías and López-Olivari, 2012; Marra et al., 2016). The T₂ treatment received an average of 185 mm per growing season, which was almost 20% less than the T₁ treatment. This treatment reached its Ψ_stem threshold (-3.5 MPa) during pit hardening after an average of 37 days without irrigation (Table 4). Reducing irrigation during pit hardening has been recommended by many authors because this phase is the least sensitive to water deficit (Goldhamer, 1999; Alegre et al., 2002; Gómez del Campo and García, 2013). Also, the application of RDI during this phase could allow for considerable water savings because pit hardening generally coincides with high atmospheric demands for water vapor (Goldhamer, 1999). In our study, the pit hardening period (maximum rate) was reached during January, which coincides with the time of the year observed in previous seasons (López-Olivari et al., 2016). The T₃ and T₄ treatments received an average of 163 and 141 mm of irrigation per growing season, respectively, and reached their Ψ_stem thresholds post-pit hardening (post-maximum rate) after 56 and 72 days without irrigation. These thresholds (-5.0 MPa in T₃ and -6.0 MPa in T₄) suggest that the trees in the T₃ and T₄ treatments were severely stressed during the experiment (Moriana et al., 2002). Once irrigation was restored in the T₂, T₃, and T₄ treatments, their Ψ_stem returned to values similar to those of the T₁ treatment as has been observed by other authors (Dell’Amico et al., 2012; Agüero et al., 2016).

The orchard was severely pruned during the spring of the 2012–2013 season, which likely decreased the daily crop water requirements due to reduced leaf area per tree. This lower demand could explain the extended period without irrigation in the T₂ and T₃ treatments before reaching their Ψ_stem threshold in 2012–2013 (50 and 71 days in T₂ and T₃, respectively). In the case of the T₄ treatment, the combination of spring pruning and rainfall in December (23.9 mm) resulted in the Ψ_stem threshold not being reached in the 2012–2013 season.

Despite receiving almost 20% less irrigation, fruit yield of the T₂ treatment (10,917 kg ha^-1) was not statistically lower than that of the T₁ treatment (11,984 kg ha^-1) for the four growing seasons. This suggests that this irrigation cut-off strategy could be applied in commercial orchards without affecting yield. In contrast, the fruit size was reduced by 15% (T₂), although maximum water stress occurred after the vast majority of endocarp (pit) and mesocarp (pulp) cells were formed (Hammami et al., 2011). This reduction can be explained because fruit expansion requires an adequate flow of water to the fruit and sufficient turgor to drive in cell enlargement (Dell’Amico et al., 2012). These results are in accordance with Marra et al. (2016), who suggested maintaining Ψ_stem values between -3.5 and -2.5 MPa to get an optimal-moderate yield in olive cv. Arbequina.

Fruit yields for T₃ and T₄ treatments were 33 and 39% less than T₁, respectively. These reductions were mainly due to a smaller crop load and lower fruit weight as has been reported by Fernandes-Silva et al. (2010) for severe water stress conditions. Crop load was reduced by 10–13% in T₃ and T₄ treatments, respectively, and fruit weight was reduced by 25–30%. This decrease in yield and their components can be explained by limitations to photosynthesis which are controlled by water stress (Angelopoulos et al., 1996; Flexas and Medrano, 2002). Indeed, the immediate response of plants to water stress is to limit leaf transpiration in order to reduce water loss through stomatal closure (Fernández et al., 1997). However, this also causes reduced CO₂ diffusion into the leaf, thereby limiting carbon assimilation (Angelopoulos et al., 1996; Ennajeh et al., 2008).

Moreover, relative fruit yield showed a linear response with the minimum Ψ_stem during the seasons, despite that there was a high dispersion of data (Figure 4). However, an integrated plant water status is more appropriated for evaluating the effect of water deficit on fruit yield. Therefore, the S_Ψ was related to fruit yield and had a high linear correlation (Figure 5). These results coincide with those reported by Moriana et al. (2012) in olive trees cv. Cornicabra. However, in our results, this relationship was strongly influenced by fruit load. Thus, in the “on” years, the relationship was higher and presented a greater slope than in “off” years.

Fruit yields of all treatments followed a biannual (alternate bearing) pattern with an “on” year in 2010–2011 and an “off” year in 2011–2012. Unfortunately, the pruning done in the 2012–2013 season prevented the assessment of alternate bearing in the last two seasons. Such a pattern is common in most olive cultivars (Lavee, 2007). This often occurs because in the “on” year there is low shoot growth, which leads to few potentially reproductive buds for the next year; when crop load is high there is inhibition of floral induction (Dag et al., 2010; Fernández et al., 2015). Consequently, the year-to-year variations in yield are directly related with fruit number per tree of each year (Martín-Vertedor et al., 2011; Trentacoste et al., 2015). Additionally, fruit size and fruit weight were lower in the seasons with higher crop loads. These results are explained by the close relationship between yield components and the number of olives per tree (Iniesta et al., 2009; Martín-Vertedor et al., 2011).

Water deficit treatments did not lead to any changes in total oil content (%). Therefore, the reductions in oil yield in the T₃ and T₄ treatments were related to crop load and fruit weight, rather than to total oil content itself. Total oil content on a dry weight basis appears to be fairly insensitive to water deficit (Patumi et al., 2002; Moriana et al., 2003; Tognetti et al., 2007; Iniesta et al., 2009; García et al., 2013). However, Trentacoste et al. (2015) found a significant reduction of 3.1% in total oil content using a -2.5 MPa irrigation threshold in cv. Frantoio. The lack of response of oil content (%) in many of these studies is likely a function of water deficit being implemented during the pit hardening period, rather than later when most oil accumulation occurs. Moreover, total oil content was significantly lower in the 2010–2011 season compared to other seasons. In this case, air temperature (minimum and maximum) and maturity index (MI) were lower during April 2011 (40 days before harvest) than in other years. In this case, the MI were 1.56, 1.69, 1.56, and 1.11 for 2010–2011, 2011–2012, 2012–2013, and 2013–2014, respectively. These results coincide with those observed by Motilva et al. (2000) and Grattan et al. (2006) who suggest that total oil content is highly related to MI.

Despite T₂ receiving almost 20% less irrigation, no significant differences in WP_f and WP_o were found between the T₁ and T₂ treatments. In contrast, WP decreased for both fruit and oil yield under severe water stress in the T₃ and T₄ treatments. Thus, the WP responses would be explained by the difference in the intensity of water deficit treatments. The decrease in WP_f and WP_o under severe stress is consistent with the results of Fernandes-Silva et al. (2010) in olive cv Cobrançosa where water potential also reached around -6.0 MPa. Some studies have reported an increase in WP_f under moderate stress conditions with Ψ_stem values similar to those observed with the T₂ treatment (Marra et al., 2016). It may be that an intermediate Ψ_stem threshold between our T₂ and the T₃ and T₄ treatments may have led to a significant increase in WP_f and WP_o.

Results obtained in the present study over four growing seasons showed that yields were not affected when irrigation cut-off was applied from fruit set until reaching a threshold level of -3.5 MPa (T₂ treatment) around massive pit hardening compared to T₁ (100% ETc). This provides evidence that this period is not overly sensitive to moderate water stress. However, yield and its components were severely affected when using Ψ_stem thresholds of -5.0 (T₃) and -6.0 (T₄) MPa. Also, the total oil content (%) and pulp/pit ratio were not affected by the different irrigation cut-off strategies. Moreover, the fruit and oil water productivities were significantly greater in T₂ compared to T₃ and T₄ treatments.

In summary, these results suggest that the T₂ irrigation cut-off strategy would be the most appropriate, because this treatment maintained fruit and oil yield, saving 20% of the total water applied. These results suggest that this strategy (T₂) is a viable strategy to be implemented in high-density olive orchards in climates similar to the one where this research was done.

Conceived and designed the experiments: SO-F and LA-O. Performed the evaluations: LA-O. Analyzed the data: SO-F, LA-O, PS, and JR. Wrote the paper: LA-O, SO-F, PS, and JR. Implemented reviewers comments: SO-F, LA-O, PS, and JR.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.