The European Union (EU) is the top world olive oil and table olives producer, being the main producer areas located in the Mediterranean basin. Spain, Italy and Greece hold more than 95% of the total EU olive oil production during 2015/16 harvesting season (IOOC, 2015). In those countries, olive growing has an outstanding economic and social importance (MAPAMA, 2016), playing an important role as economic activity in a large number of rural areas. Currently, the decrease of competitiveness in olive sector threatens the survival of olive farms and mills (Pomarici & Vecchio, 2013), that needs to improve farm efficiency through modernisation, shared machinery management or outsourcing labours (Vilar-Hernández et al., 2011). In the last decades, olive orchards management evolved towards more intense planting densities and use of high level of mechanization (Novello et al., 2014); however, there is still margin for improvement in the farming techniques (Carmona-Torres et al., 2014).

Manual fruit harvesting is unavoidable for many fresh fruit and vegetables, but it suffer low working capacity and uncertain labour availability, although orchard layout, dwarf trees and picking aids can improve manual harvesting performance (Sanders, 2005). Furthermore, manual fruit harvesting may contribute to the development of ailments in the worker musculoskeletal system (Mlotek et al., 2015) as well as hand held shaker combs and branch shakers (Çakmak et al., 2011). Nevertheless, in a large number of olive orchards, harvesting is currently performed using different mechanical devices such as trunk shakers, hand held devices or integral harvesters. In large olive growing areas there are high labour requirements during harvesting operation, mainly for traditional olive orchards in which effective field capacity for harvesting operation are often poor (Famiani et al., 2014; Castillo-Ruiz et al., 2015a). Current olive harvesting systems achieve harvest efficiency ranging from 90%, achieved by trunk shakers in favourable conditions, to 80%, achieved by lateral canopy shakers, while manual harvesting can reach 98% of total harvest efficiency (Sola-Guirado et al., 2014). However, trunk shaker performance depends on several external conditions such as fruit ripeness (Blanco-Roldán et al., 2009), tree structure (Castro-Garcia et al., 2008) or machine features (Castro-Garcia et al., 2015). Furthermore, it would be desirable that commercial mass harvesting systems for olives will achieve harvesting efficiencies above 90% (D’Agostino et al., 2008), without using manual or hand held systems to remove left fruit. In fact, the breakeven point for olive harvesting efficiency is considered 85% (Farinelli et al., 2012a) due to limitations of commercial harvesters in standard harvesting conditions. Nonetheless, straddle canopy shakers usually achieve harvesting efficiency values over 95% (Farinelli & Tombesi, 2015), although these systems have different harvesting efficiency depending on the cultivar (Vivaldi et al., 2015), canopy volume and training system (Tombesi & Farinelli, 2014).

Although harvesting at early ripening stages is becoming a current technique to enhance phenol content in fruit for premium olive oil, at early harvesting season, it is more difficult to achieve high harvesting efficiency (Blanco-Roldán et al., 2009). Therefore, a better understanding of the detachment process is essential to improve mechanical harvest efficiency. Up to date, fruit detachment force (FDF) is used as the main index to describe the resistance of fruit to detachment from the tree, but it measures only traction force. It was divided by fruit fresh weight (FDF/FW) to obtain a more representative index able to predict the harvesting efficiency of trunk shaker (Farinelli et al., 2012b). Furthermore, FDF/FW decreases when fruit ripeness moves forward or when abscission chemicals are applied (Sessiz & Özcan, 2006). However, up to date it is unclear how FDF, acceleration or tree architecture are involved in fruit detaching process (Tombesi et al., 2017).

Current harvesting systems cause a limited stalk twisting. In trees subject to forced vibrations produced by trunk shakers, fruit experienced stalk twisting angles under 70 °, with peak angles around 150 º (Castillo-Ruiz et al., 2016). Moreover during cherry harvesting using limb shaker, twisting has limited influence on the number of motion patterns in comparison with tilting or beam column motion patterns (Zhou et al., 2016). Furthermore, some test and simulations performed in oranges harvested by a canopy shaker determine that only 18% of FDF was applied to fruit stalk suggesting that fruit undergoes twisting and bending processes during mechanical harvesting (Savary et al., 2011).

Fruit detachment is affected by stalk geometrical properties and its behaviour during tree shaking (López-Jiménez, 1979). Motion of fruit-stem subsystem can be described with three translational and three rotational degrees of freedom, which corresponds to precession, mutation and spin of the fruit (Upadhyaya et al., 1981). Concerning stalk structure, it can be divided into three different parts from bearing branch to fruit, these parts are peduncle, rachis and pedicel. These sections may be considered as different abscission areas for olive, and fruit abscission is affected by harvesting date and cultivar, while fruit fresh weight (FW) does not show a significant effect (Castillo-Llanque & Rapoport, 2009).

The aim of the present work was to determine FDF evolution under different stalk twisting angles along the ripening process in different Spanish and Italian cultivars. Cultivar influence on stalk susceptibility to torsion strain was also assessed in order to give data for present and future harvesting systems.

FDF was significantly reduced when stalk twisting was applied before pulling the fruit out. Differences were kept along fruit ripeness process, although stalk sensitivity to torsion strains varied, depending on the sampling date. Moreover, significant differences ( p < 0.05) were commonly found over 180 º of stalk twisting, and when ripening process was advanced, fruit were often detached only by applying a stalk twisting over 360 º (Figs. 3 and 4).

Generally, FDF reduction along with stalk twisting was lower at advanced stages of ripening, as a consequence, percentage of detached fruit without traction increased. Olive stalk susceptibility to stalk twisting, demonstrated that stalk torsion strain could play an important role for olive harvesting, when it caused an important percentage of fruit detachments by itself, particularly at the end of the ripening process (Table 1).

Stalk length could explain FDF behaviour at different stalk twisting angles. Stalk length showed significant differences ( p < 0.05) for the four tested cultivars. These cultivars could be grouped in two categories: short stalk cultivars, such as Arbequina and Leccino, and long stalk cultivars, such as Frantoio and Maurino, although no significant differences were found between Leccino and Maurino (Fig. 5).

Fresh FW and oil accumulation reached a saturation level at the end of the ripening process, as particularly evident in Leccino and Maurino cultivars. This point indicated the optimal harvesting period in terms of incomes and harvesting ease. Trees were harvested in mid (Arbequina and Frantoio) and end November when optimal maturity, established on the base of oil content and FDF, was reached (Fig. 5).

Sampling process was carried out along the whole oil accumulation process, comprising an important part of the fruit growing season with maturation indexes from 0 to 6 within the Jaen maturation index (Fig. 6). Oil content on a dry weight basis was an indicator for optimal harvesting date regarding economic yield, while FW could be used as a predictor of fruit harvesting ease. Fig. 6 shows that all cultivars provided a significant quadratic trend for maturation Jaen index ( R ²=0.84; p < 0.01) (Eq. [2]) for the sampled period, while fruit firmness provided significant quadratic trend ( R ²=0.8; p < 0.01) (Eq. [3]).

Maturation Jaen index= -0.00124 · DOY2 + 0.81 · DOY - 126.6 [2]

Fruit firmness (N)= 0.0016 · DOY2 - DOY + 157.6 [3]

All studied cultivars showed high susceptibility of FDF and FDF/FW ratio to stalk twisting, while some cultivars such as Arbequina, Leccino and Maurino varied FDF at a lesser extent along ripening process than along stalk twisting. Fruit detachment became easier along ripening process because of FDF reduction, although this reduction was usually greater only by applying a stalk twisting angle over 360º. However, all cultivars did not show the same pattern: Arbequina, Leccino and Maurino decreased more regularly along ripening, while Frantoio was more irregular. Furthermore, FDF was consistently reduced at wider stalk twisting angles in all cultivars suggesting a possible facilitation of harvesting process mainly for early harvesting. FW played also a role along the maturation process, softening FDF changes in frequently sampled cultivars as Arbequina and Frantoio, while in less frequently sampled cultivars, FDF/FW ratio provided higher differences between different stalk twisting angles than along ripeness process (Figs. 3, 4 and 8).

FDF decreased when it was measured after applying stalk twisting angles. FDF reduction was greater when stalk twisting was higher, which could play an important role in fruit detachment process. Although in other crops twisting movement pattern has limited importance (Zhou et al., 2016), in olive inertial and bending forces can be key factors in fruit detachment process (Tsatsarelis, 1987). Therefore, current and future harvesters could take advantage of torsion strain at stalk level, due to the effect on FDF decrease. This could be particularly important for increasing harvesting efficiency in early harvesting, when trunk shaker causes a lower percentage of detached fruit than in late harvesting (Blanco-Roldán et al., 2009).

Stalk twisting had different effect on FDF depending on cultivar: Leccino usually had higher FDF for all twisting angles than Maurino, while Arbequina and Frantoio had a more erratic behaviour. In early harvesting, Frantoio was roughly more difficult to detach as long as Arbequina was more difficult to detach during the final part of the ripening process. Those differences can be related to stalk length, physiological aspects or water stress. Furthermore, FDF depends strongly on stalk diameter (Lavee et al., 1982) varying between different olive cultivars (Farinelli et al., 2012b). Stalk length may affect harvesting performance due to vibration transmission from fruit bearing branches to stalk-to-attachment point or to stalk-to-fruit attachment point.

Provided data support the hypothesis that bending forces collaborate along with traction and inertial forces in fruit detachment process, as reported fruit species: in oranges, FDF decreases when the pulling direction forms a greater angle with the pistil-calyx axe (Torregrosa et al., 2014). Furthermore, in manual apple picking bend and pull forces are combined to reduce detachment energy as compared to the application of sole pulling force (Li et al., 2016). Several fruit twisting can also facilitate fruit picking (Chiu et al., 2013), mainly for table olives, which are highly susceptible to bruising.

Harvesting efficiency depends on several factors such as shaking frequency (Leone et al., 2015; Tombesi et al., 2017), abscission agent application (Sessiz & Özcan, 2006), tree features (Farinelli et al., 2012a), time, and harvesting date (Castro-Garcia et al., 2015). Different shaking technologies and vibration patterns could be applied to olive harvesting with similar final results (Sola-Guirado et al., 2014). Different harvesting systems have variable harvesting efficiency within the canopy with possible effect on the oil quality considering that fruit with the highest oil yield and oil quality are located in the outermost part of the canopy (Castillo-Ruiz et al., 2015b). Finally, future machines should match bending, torsion and traction forces with inertial ones to accomplish a quick and effective fruit detachment that reduces required hand labour, which is currently used to assist trunk shaker harvesting with long poles or hand held devices.

Fresh FW increased along the maturation process due to tissue growth and oil accumulation (Farinelli et al., 2002). FW played an important role during olive harvesting process, FDF/FW affects harvesting efficiency (Farinelli et al., 2012b) mainly due to stalk bending forces, inertial phenomena and fatigue (Tsatsarelis, 1987). During early harvesting, high FDF reduces harvesting efficiency, thus, it is necessary to increase shaking time to achieve high removal efficiency (Blanco-Roldán et al., 2009). But prolonged tree shaking can cause bark damages in particular when the tree is still vegetative (Gurusinghe & Shackel, 1995).

All tested cultivars provided an FDF value under 3 N when the stalk twisting was over 180º except Leccino, which reached the same values when stalk twisting was over 360º. Therefore, FDF measurement combining pulling and twisting forces could also be a useful index to predict both, harvesting efficiency and maximum oil content on a dry mass basis, determining the optimal harvesting time (Portarena et al., 2015).

Results showed a quadratic trend for maturation Jaen index and fruit firmness along the ripening process. However previous research, based on non-destructive methods, reports that olive skin colour followed a quadratic trend for Arbequina and Picual cultivars while firmness follows a linear trend (Garcia & Yousfi, 2005). Other non-destructive methods to measure olive ripeness could be based on near infrared spectroscopy (Gracia & León, 2011) considering that fruit firmness is an important variable which influence fruit damages during mechanical harvesting (Tombesi et al., 2011). Furthermore, for straddle harvesters based on canopy shaking, harvesting efficiency could be predicted using firmness, colorimetric and pigmentation indexes (Camposeo et al., 2013).

Currently, for olive and other fruit crops, fruit ripening influence on harvesting process is measured by FDF (Zipori et al., 2014) and FDF/FW ratio (Polat et al., 2011) applying only traction forces to the fruit stalk. Each harvesting machine can cause a different stalk twisting angle depending on the machine-tree interactions (Tombesi et al., 2017). Once mean twisting angle is known, it would be possible to get more reliable estimation of expected harvesting efficiency on the base of FDF and FDF/FW ratio at the expected stalk twisting interval. This methodology could be useful to predict a mechanical harvest efficiency over 85% particularly for early harvesting, considering that it takes place when FDF/FW ratio goes under 2.3 (Farinelli et al., 2012b). Further research is required to explain how other forces such as bending or inertial forces influence fruit detachment in olives and other crops as well as how climatic conditions affect FDF and FDF/FW evolution.

In summary, olive stalk resistance against pulling force was reduced when stalk twisting was applied in some olive cultivars. Since earliest sampling dates, all cultivars showed FDF under 3 N at stalk twisting angles over 180 º, except for Leccino, that required wider angles (up to 360º). At the same time, FDF and FDF/FW ratio were reduced along the ripening process, although FDF/FW ratio provided less variability data and also took into account inertial forces during harvesting process. Combined pulling and twisting forces provide a better estimation of the real fruit susceptibility to detachment, closer to real harvesting. Jaen ripeness index, fruit firmness and oil content on a dry matter basis should be used along with FDF at different twisting angles, as predictors for optimal harvesting period in order to achieve the highest profitability for olive growers. Finally, present and future harvesting systems should take advantage of stalk susceptibility against torsion or bending strain to achieve higher harvesting efficiency.