A 3-year experiment was established in two commercial olive orchards located in Valdepeñas (Ciudad Real), central Spain (38º50’ N, 3º19’ W). The distance between both orchards was about 300 m and the soil type was similar (Typic Rhodoxeralf). In the cv. Arbequina orchard, 11-year-old olive trees were trained in hedgerow at a spacing of 4 x 1.5 m (1667 olives/ha). The soil was sandy loam at the first 0.3 m and clay below this depth. The pH was 7, CaCO₃ was < 0.5%, and organic matter was 1.3% in the first 0.15 m of the soil and < 0.6% below this depth. P and K contents were 118 mg·kg⁻¹and 484 mg·kg⁻¹ at the first 0.3 m, and < 1.4 mg·kg⁻¹ and 247 mg·kg⁻¹ below this depth, respectively. In the cv. Picual orchard, 12-year-old olive trees were vase-trained at a spacing of 8 x 6 m (208 olives/ha). The soil was sandy loam at the first 0.18 m and clay below this depth. The pH was 7.1, CaCO₃ was < 0.5% at the first 0.48 m and 1.3% below, and organic matter was 1.0% at the first 0.18 m of the soil and < 0.6% below. Available P was 38 mg·kg⁻¹ the first 0.18 m and < 1.5 mg·kg⁻¹ below, whereas exchangeable K was 760.5 mg·kg⁻¹ in all profiles. Olive trees were pruned annually following their trained systems. Weeds were controlled using pre- and post-emergence herbicides. The climate in the area is typically Mediterranean, with mild rainy springs and hot, dry summers. During the experiment, a weather station 15 km from the site registered temperature, rainfall, wind speed and direction, humidity, and global radiation and calculated ETo by the Penman-Monteith method (Allen et al. 1998) from these climatic variables. The mean annual temperature, ETo and rainfall (11-year average) were 13.8 ºC, 1328 mm and 367 mm, respectively.

Both orchards were deficit irrigated by a subsoil drip system of 3.5 L·h⁻¹ every 1.5 m in ‘Arbequina’ and 2 m in ‘Picual’. Irrigation doses were applied by the grower according to maximum doses allowed by Guadiana River Basin Authorities each year. So, ‘Arbequina’ orchard olive trees received 78, 84 and 121 mm in the first, second and third year, respectively, and ‘Picual’ orchard olives received 15, 21, and 29 mm, respectively.

During the three years of the experiment, four N-fertilization treatments were applied in the ‘Arbequina’ orchard (N, N+i, 2N and 2 N+i) and two N-treatments were applied in the ‘Picual’ orchard (N, N+i). Nitrogen was applied as ammonium sulphate (21% NH₄⁺) alone or with 0.8% nitrification inhibitor DMPP (3.4-dimethylpyrazolphosphate). In all cases nitrogen was applied in spring to the soil under the canopy projection, 50% of the dose was applied in April and the other 50% in May. In addition, all trees received a foliar fertilization applied by the grower with N, P and K annual rates of 0.4, 0.4 and 1.7 kg N·ha⁻¹in ‘Arbequina’ and 0.4, 0.4 and 0.5 kg N·ha⁻¹in ‘Picual’; 0.1, 0.1 and 0.9 kg P₂O₅·ha⁻¹; and 0.1, 3.1 and 1.9 kg K₂O·ha⁻¹ on both orchards, respectively. Total doses of N annually applied including the grower fertilization are summarized in Table 1. In the first year of the experiment, the doses applied corresponded to doses usually applied by growers. Because no tree response to fertilization was obtained, during the second and third years N-fertilization was reduced following regional authorities’ advice in order to reduce nitrate pollution of the groundwater in this area.

N-fertilization treatments and non-treated control trees (CON) were randomized in a complete four-block design. The experimental plot consisted of three trees surrounded by 2 guard trees.

Leaf nutrient concentration was determined each year from 100 mature leaves sampled in July from the middle of the basal portion of non-bearing, current season shoots around the trees and at about 1.5 m above ground. Leaves were collected from each plot in three blocks per treatment and the control. Once in the laboratory, the leaves were rinsed in de-ionized water, dried at 60 ºC, and mineralized at 450 ºC. After that, the ashes were digested with acid and the extract was measured by a ICP-OES spectrometer (Perkin Elmer, Massachusetts, USA) to determine P, K and Ca, Mg, Na, Fe, Mn, Cu, Zn, Mo, Sulfates and B. Nitrogen was determined with a LECO TruSpec^®NNitrogen Protein Analyzer System (LECO, Michigan, USA. The methodology was described by Fernández-Escobar et al. (2009).

In March the trunk perimeter of 3 control trees per treatment and block were measured at 35 cm above the soil surface. In each control tree, 3 shoots were tagged on the S-side. Vegetative growth was evaluated by measuring shoot length and number of node increments from budburst (March) to harvest (end of October in ‘Arbequina’ and end of November in ‘Picual’). In these tagged shoots reproductive components were measured. The number of inflorescence and the number of fruits per shoot were counted in May and at harvest, respectively. These observations were used to calculate the percentage of buds that developed inflorescence (buds initiated) and percentage of inflorescence at the onset of fruit, at least one flower (fertile inflorescence).

Two trees per treatment and block were harvested individually (‘Arbequina’ early to mid-November and ‘Picual’ at the end of November) and fruit production was weighed each year. From harvested fruits, six subsamples in ‘Arbequina’ and ‘Picual’, close to 25 g, were weighed fresh and again after drying at 105 ºC. In these subsample fruits were counted to determine fruit weight. A maturity index was determined based on the color of skin and pulp (Ferreira, 1979). Oil content was determined by nuclear magnetic resonance (MiniSpec MQ-10, Bruker, Madison, USA) using the method described by Del Rio & Romero (1999). Oil content was calculated on fresh and dry bases. Fruit water content was calculated as a difference between fruit fresh and dry weights.

Another subsample of 2.5 kg was used for oil extraction with the Abencor system (Commercial Abengoa S.A., Seville, Spain) in 2009 and 2010. This unit, consisting of three basic elements, a hammer mill, a thermo-beater, and a pulp centrifuge, simulates the industrial process of virgin olive oil production on a laboratory scale (Martínez et al., 1975). Samples were crushed in the hammer mill at 3000 rpm. The resulting olive paste was placed in stainless steel 1-L containers and malaxated for 45 min in the thermo-beater at 26 ºC, using four stainless steel cross blades at 54.5 rpm (radius 53 mm). After that, the paste was centrifuged in a pulp centrifuge for 1 min at 3,500 rpm (radius 100 mm) to separate the liquid phase (oil and waste water) from the solid waste. Then, the oil was decanted into graduated tubes until complete separation of the water and oil phases was obtained. Oil volume was determined after decantation and the extractability index was calculated as the percentage of olive oil extracted from the total oil content of the fruit (on a fresh matter basis) considering 0.916 kg·L⁻¹, the olive oil density at ambient temperature. After measurement, the oil was filtered through filter paper and stored in a N₂ atmosphere at -20 ºC until analysis.

Free acidity, peroxide index value, and coefficients of specific extinction at 232 and 270 nm (K₂₃₂ and K₂₇₀) were evaluated in the last two years according to the Regulation EEC/2568/91. An automated Methrom Rancimat 679 apparatus (Methrom Co., Basel, Switzerland) was used to determine the oxidative stability with a 2.5 g oil sample warmed to 98 ºC and an air flow of 10 L·h⁻¹. Results were expressed as induction time in hours (Gutiérrez, 1989). The composition of fatty acids was determined by gas chromatography in a Perkin-Elmer Autosystem (CT, USA). The fatty acids (carbon number:unsaturations) analyzed were myristic (14:0), palmitic (16:0), palmitoleic (16:1), margaric (17:0), margaroleic (17:1), stearic (18:0), oleic (18:1), linoleic (18:2), linolenic (18:3), arachidic (20:0), gadoleic (20:1) and behenic (22:0). From the fatty acid composition, different variables were determined: saturated fatty acids (SAFA), including fatty acids without unsaturations; unsaturated fatty acids (UNFA), including fatty acids with one, two or three unsaturations; monounsaturated fatty acids (MUFA), including fatty acids with only one unsaturation; and polyunsaturated fatty acids (PUFA), including fatty acids with two or three unsaturations. The phenolic fraction of the oil samples was isolated by solid-phase extraction and analyzed by reversed-phase HPLC using a diode array UV detector (Mateos et al. 2001). The quantification of phenolic compounds (except ferulic acid) was carried out at 280 nm using p-hydroxyphenylacetic acid as an internal standard, whereas that of flavones and ferulic acid was made at 335nm using o-coumaric acid as an internal standard. The results were expressed in ppm.

Analyses of variance were performed on the data using MSTAT-C (University of Michigan, USA) to compare the effect of treatments. Least significant differences (P ≤ 0.05) tests (protected LSD) were used to separate the means of all parameters evaluated between treatments. The error term for LSD was calculated from the residual of the ANOVA. All percentage values were transformed using the arcs of the square root before analysis.

Weather data of the experimental periods are represented in Figure1. Annual rainfall in 2008, 2009 and 2010 was 419, 392 and 652 mm, respectively, and annual ETo was 1250, 1305 and 1198 mm, respectively. Annual rainfall distribution was variable particularly from January to May when 236, 100 and 317 mm fell in the 1^st, 2^nd and 3^rd years, which meant 56, 26 and 49% of the total annual rainfall, respectively. In this period, ETo was 393, 433 and 395 mm, respectively. June and July were very dry with 5, 8 and 7% of total annual rainfall, and ETo was 397, 409 and 375 mm, respectively in 2008, 2009 and 2010. From August to October rainfall was 114, 101 and 89 mm, which meant 27, 26 and 13% of the total annual rainfall respectively, and ETo was 398, 379 and 365 mm, respectively. The maximum temperature in summer was measured in July at 39 ºC, 39.6 ºC and 39.7 ºC, respectively. November and December were the coldest months with absolute minimum temperatures of -5.6 ºC, -12.1 ºC and -8.4 ºC, respectively.

In the ‘Arbequina’ orchard foliar N-concentration in CON varied between 1.5% in 2009 and 1.9% in 2010 (Table 2). In the last year N-fertilization treatments increased foliar nitrogen concentration more significantly than CON and presented a N-leaf composition higher than 2%. In 2N and 2N+i, P- and K-leaf levels were significantly lower than the rest of the treatments and CON. A reduction in the P-leaf level in the most N-fertilized treatments was also observed on the average of the three years of experiment.

In the ‘Picual’ orchard, N-leaf levels in CON varied between 1.7% in the 2^nd year and 2% in 2010 (Table 2). In this orchard N-fertilization treatments did not significantly modify the foliar N-P-K concentration.

In the ‘Arbequina’ orchard, N and N+i fertilization treatments increased vegetative growth significantly with respect to CON in the last two years (Table 3). These treatments increased trunk perimeter significantly in 2009 and shoot elongation and number of nodes developed in 2010. In this last year, 2N significantly increased shoot elongation with respect to CON by 51%. In 2008, N and N+i increased the percentage of buds initiated. Considering the whole period, the nitrification inhibitor applied in 2N+i significantly developed more nodes per shoot than CON by 39%, and in N+i the percentage of buds initiated was significantly increased regarding the other treatments. Fertile inflorescence was not significantly affected by the fertilization treatments. All parameters were significantly affected by the year. Vegetative growth and the percentage of bud burst and fertile inflorescence were higher in 2010. However, in 2009 the percentage of buds initiated was higher.

In the ‘Picual’ orchard, N fertilization did not have any significant effect on growth variables or reproductive components (Table 4). Year had a significant effect on reproductive components and vegetative growth, except for trunk perimeter. Shoot elongation and number of nodes were higher in 2008. Nevertheless, the percentages of bud burst and buds initiated were lower.

In the ‘Arbequina’ orchard fruits were harvested with a skin color between deep green and yellow-green (average maturity index MI=0.5) (Table 5). Production varied significantly with year. In 2010, olive and oil production were higher.

In ‘Arbequina’ fertilization treatments only produced significant differences the last year in fruit characteristics (dry weight and oil per fruit) but not in olive or oil production (Table 5). Year significantly determined the effect of N-treatments although fruits were harvested at the same maturity index. The interaction between year and fertilization was not significant in these evaluated variables. The treatments that received the higher N-doses (2N and 2N+i) presented fruits with 14% lower dry weights than CON. Furthermore, fruits from 2N had 24% lower oil content. In the whole period, N+i, 2N and 2N+i had 20% more fruits per tree than CON; however, there were no significant differences in oil production.

In the ‘Picual’ orchard, olives were harvested with a skin color between pink or purple in more than half of the fruit and black skin with white pulp (MI between 3.1 and 4) (Table 6). In 2010 production was higher than the rest of the seasons. In the first year, N+i increased fruit dry weight, water content and oil per fruit by 52%, 3.3% and 73% regarding CON, respectively, and by 30%, 3.3% and 43% regarding N, respectively. There were no significant differences between treatments for other seasons or variables.

Oil quality was evaluated in the two latest years by measuring acidity, peroxide index, K₂₃₂ and K₂₇₀ values, and fatty acid and phenolic compositions. In both orchards and treatments, fatty acids such as linoleic, myristic, margaric, margaroleic, araquic, gadoleic and behenic presented very low concentrations (< 0.8%) and were not affected by treatments or year (data not shown). Only the phenols with high concentrations or those showing significant differences between treatments have been studied.

In the ‘Arbequina’ orchard, year had a significant effect on oil quality parameters. In the oil obtained from CON fruits free acidity reached very low values (0.1-0.2% oleic acid), peroxide value was lower than 6 meqO₂·kg⁻¹, K₂₃₂ and K₂₇₀ were lower than 1.80 and 0.20, respectively and oxidative stability was lower than 36 hours (data not shown). Year also showed a significant effect on fatty acid and phenolic compositions (Table7). During the last year oleic acid concentration, C18:1/C18:2, MUFA/PUFA and UNFA/SAFA were higher than the previous one. However, the phenolic content was higher in 2009 than in 2010. In CON, total phenols, total O-diphenols and total secoiridoids in 2010 were respectively 74, 77 and 80% lower than those obtained in 2009. In all seasons, there was a positive linear relationship between stability and total phenols (y = 42.4x – 679.8, R² = 0.98, n=15), total O-diphenols (y = 28.3x – 482.4, R² = 0.98, n=15) and total secoiridoids (y = 41.9x – 721.7, R² = 0.98, n=15).

In this orchard, N-fertilization did not produce significant differences in oil quality variables or fatty acid composition individually considered or when ratios were analyzed. However, N-fertilization with and without nitrification inhibitor negatively affected the oil’s phenolic composition (Table 7). During the last year O-diphenols were reduced in all N-fertilization treatments and N-treatments with nitrification inhibitor presented lower total phenols, total O-diphenols and total secoiridoids than CON.

In the ‘Picual’ orchard, year had a significant effect on oil quality parameter variables except for oxidative stability, which was an average of 61.5 hours (data not shown). In this orchard, the relationship between stability and phenol content was not significant. Free acidity in the oil from CON reached very low values, lower than 0.3% oleic acid, peroxide value was lower than 4 meqO₂·kg⁻¹, and the values of K₂₃₂ and K₂₇₀ were lower than 1.7 and 0.2, respectively (data not shown). Year also had a significant effect on fatty acid composition and in the last year oleic acid concentration was higher than 2009 (Table 8). Regarding phenolic compounds there were no differences between years.

In this orchard, N-fertilization did not produce significant differences in any variables of oil quality, in phenolic composition or in the fatty acid composition individually considered. However, in the last year N+i increased significantly the UNFA/SAFA ratio regarding CON and N by 2%.