Eleven tests were conducted on five commercial orange (Citrus sinensis (L.) Osb) and mandarin groves, including clementines (Citrus clementina Hort. ex Tan.) and hybrids (Citrus clementina Hort. ex Tan. × Citrus tangerina Hort. ex Tan.) during the seasons 2008-09, 2009-10 and 2010-11. The characteristics of each grove are shown in Table 1. These groves produced from early (‘Marisol’) to late hybrid mandarins (‘Fortune’), that are harvested in autumn and winter, thus covering the part of the harvest season in which there are few data from other countries.

Five treatments were carried out in each test: one control (water) and four different doses of ethephon (Ethrel 48, Numarf España, S.A., Barcelona, Spain) resulting from the combination of (i) two concentrations (600 and 1200 ppm) and (ii) two spray volumes, one higher, which was defined as the volume of liquid until the runoff point (V_H), and one lower, which was defined as a 40% reduction of the higher volume (V_L). These volumes varied according to the volume of vegetation in the canopy of the trees (V_Veg) of each variety and are shown in Table 2. An adjuvant (Mojante Inagra, Sipcam Inagra, S.A., Valencia, Spain) was added to the treatments at 0.05% to facilitate distribution of the product, as it has been indicated in the literature (Burns et al., 1999, 2006a,b; Kender et al., 2000; Pozo & Burns, 2009). The pH of the mixture was around 6.8-7 in all the tests. Trials were made in a completely randomized experimental design. The experimental unit was one tree and each treatment was repeated five times, with a total of 25 trees per test.

The treatments were carried out with a hydraulic handgun sprayer, with a working pressure of 3 MPa and a cone angle of 30°. The higher water volume was applied using a ceramic conical nozzle of 1.2 mm diameter, whereas the lower water volume was applied by means of the same type of nozzle but with 1 mm diameter.

Between 6 and 12 days after applying ethephon (Table 2), all the trees were harvested with an orbital trunk shaker (Topavi, model: vibrator support arm, Maquinaria Garrido S.L., Autol, La Rioja, Spain) equipped with a three-point grip system (Fig. 1). Table 2 shows the operating characteristics (frequency, F; amplitude, A; and duration of vibration, t) of the shaker in each test. The frequency in the different tests ranged between 14.1 Hz and 15.5 Hz and the amplitude between 15-35 mm (Ortiz & Torregrosa, 2013). The duration of vibration was 5 s in all trials except for the first trial. Shaking was applied in one (5 s) or in two times (3 s+2 s), except in the first trial, in which trees were shacked 10 s (5 s +5 s). Previous work (Torregrosa et al., 2009) demonstrated that these shaking patterns have little importance in the results, since fruit and leaves are detached in the first 3 s. Frequency and duration of vibration were measured with a triaxial accelerometer placed on the tree trunk, near the shaker arm, registering the whole duration of the shakes with a digital oscilloscope at a frequency of 585 Hz. Amplitude was measured with video records at 300 frames per second. During shaking both fruit and leaves fell onto canvases that were arranged under each tree to catch them.

The dates when tests began were defined by the commercial demand for the fruit and the weather. Weather conditions (temperature, mean relative humidity and rainfall) from 15 days before the application until harvesting are shown in Table 3. Between the application of ethephon and the harvest with the trunk shaker there was very little rainfall. The small amount of rain fallen in the early days of the season before the ‘Marisol’ mandarins were harvested had no influence on the effect of ethephon, because the 8 h needed for the plant to absorb the product had already elapsed when the rain started (Wilson et al., 1977, 1981).

To determine the state of the fruit before applying ethephon, five fruits were picked at random from control trees, leaving a minimum stalk length of 2 cm. The fruit retention force (FRF) and citrus colour index (CCI) were measured for each fruit. Total soluble solids (TSS), acidity, and maturity index (MI) were measured for each juice extracted from the sample of five fruits of control trees. The FRF was measured using a digital dynamometer (Advanced Force Gauge 500 N, Mecmesin, England) by holding the stalk of the fruit horizontally in a fixed clamp and leaving 0.5 cm free until the calyx. The fruit was then pulled with the dynamometer, using a structure that allowed to pull the fruit horizontally. The colour of the peel was measured with a Minolta Colorimeter (Model CR-400/410; Japan) with the Hunter Lab coordinates (Jiménez-Cuesta et al., 1981). The CCI of each fruit was calculated as the mean of two measurements taken on the equatorial zone, one in the green side and the other in the orange one. The MI was calculated as the ratio between the soluble solids and the acidity (González-Sicilia, 1968). The concentration of soluble solids was measured with a digital refractometer (Atago model PAL-3; Atago Co., Tokyo, Japan). Acidity was determined by titrating 5 mL-aliquots of juice with a 0.1 N solution of NaOH, with an automatic titrator (Mettler Toledo T50, Rondo Tower, Switzerland). In the case of the ‘Marisol’ and ‘Fortune’ mandarins, 3 mL-aliquots of juice were titrated because, being more acidic varieties, a greater amount of NaOH was needed.

Harvesting efficiency of the shaker was measured with the following two variables: fruit detached (%) and fruit detached without calyx (%). The variable defoliation was also measured. It was not possible to measure the percentage of defoliation because we were not allowed to defoliate the trees since trials were performed in commercial orchards, so we measured the amount of leaves (kg) detached per tree.

Fruits fallen after mechanical shake were weighed with a digital dynamometer (Advanced Force Gauge 500 N, Mecmesin, England). Fruits remaining on the tree after shaking were manually harvested and weighed. The relation between the amount of fruit detached with the shaker and the total amount of fruit on the tree (harvested with the shaker plus hand-picked) was used to calculate the percentage of fruit detached by the shaker.

The percentage of fruit without calyx was obtained from a random sample of 100 fruits detached by the shaker. All leaves detached from each tree after shaking were collected and weighed to evaluate defoliation.

First, the influence of the factors Season and Orchard on the efficiency of the trunk shaker was studied. This was carried out for the trees that were not treated with ethephon (dose 0) using multifactor analysis of variance on the data concerning percentage of fruit detached and percentage of fruit without calyx. Least Square Difference (LSD) test was used for mean comparisons. In this study, the assumption of normal distribution of data was assessed using the normal probability plot of the residuals and the assumption of homoscedasticity using the Levene’s test (Levene, 1960). In all the analyses a confidence level of 95% was considered.

Second, the effect of ethephon dose on these same variables was studied, also including the influence of the factors Orchard and Season. Multiple Linear Regression (MLR) was performed to study the relationship between each of the two dependent variables (percentage of fruit detached and percentage of fruit detached without calyx) and the ethephon dose. In order to test whether these relationships were affected by the factors Orchard and Season, indicator variables were included in the regression model. An indicator variable is one that takes the value 0 or 1 to indicate the absence or presence of a categorical effect that may be expected to shift the outcome. When an indicator variable has n categories, only (n − 1) indicator variables are introduced in order to avoid multicolinearity. The category for which the indicator is not assigned is known as the base group (Suits, 1957). In the present case, the factor Orchard had 6 categories (‘Orogrande’ A, ‘Orogrande’ B, ‘Marisol’, ‘Navel Lane Late’, ‘Clemenules’ and ‘Fortune’ orchards) and ‘Orogrande’ A orchard was chosen as the base group. The factor Season had 2 categories (2009-10 and 2010-11) and 2010-11 was chosen as the base group. MLR analysis followed an iterative process in which all the experimental data were included. It started by including the ethephon dose as independent variable, the two indicator variables (Orchard and Season) and their interactions in the model. Then the variable with the highest, non-significant p-value (α > 0.05) was eliminated and the model was recalculated until all variables present in the model had significant coefficients. In all fitted models, all the assumptions of linear regression were checked. No outliers were identified.

Third, because the ‘Orogrande’ A orchard was the only one that was studied along the three seasons (2008-09, 2009-10 and 2010-11), the analysis of the effect of the season on the results of the different ethephon doses on this orchard was performed following the previous methodology. In this case, the factor Season had 3 categories (2008-09, 2009-10 and 2010-11) and 2010-11 was chosen as the base group.

Due to the fact that different orchards had different sizes and leave densities, the variable defoliation (kg leaves/tree) should not be compared among orchards. For this reason, the effect of ethephon dose was studied for each orchard. In the orchards studied along several seasons (‘Orogrande A’, ‘Orogrande B’, ‘Marisol’, ‘Navel Lane Late’), the effect of the season was also studied. MLR was performed following the above methodology. As stated before, the factor Season had 2 or 3 categories depending of the orchard (2008-09, 2009-10 and 2010-11) and 2010-11 was chosen as the base group in all cases.

Table 3 summarises the values (mean ± standard error) of the state of maturity of the fruit from the reference trees in each orchard and season at the time of harvesting (CCI, MI and FRF). In season 2010-11, fruit was greener than in the season 2009-10 in all orchards, since CCI were lower. The MI values did not vary much from one season to another for each orchard, except for ‘Navel Lane Late’ orchard which fruit had lower MI in season 2010-11 than in 2009-10. In general, MI values ranged between 11-13 in the less acid varieties (‘Orogrande’, ‘Navel Lane Late’ and ‘Clemenules’) and were lower (between 5 and 8) in the more acid ones (‘Marisol’ and ‘Fortune’). In any case, all the CCI and MI values are considered adequate for marketing in our agroclimatic conditions.

The FRF values did not vary much from one season to another in each orchard. In all the seasons of ‘Orogrande’ and ‘Clemenules’ orchards, the FRF ranged between 66 and 76 N; in ‘Marisol’ and ‘Fortune’ orchards between 36 and 50 N; and in ‘Navel Lane Late’ orchard it was between 123 and 146 N.