INVESTIGACIÓN Quality of virgin olive oil as influenced by origin area

Se ha realizado una investigación para evidenciar la importancia y la Influencia que la zona de origen tiene sobre las variables analíticas del aceite de oliva virgen. Han sido seleccionadas cinco zonas geográficas de Italia: (i) una cerca de la Facultad de Agraria de la Universidad de Perusa (UNI-PG); (ii) una cerca del Instituto para la Olivicultura Consejo Nacional de las Investigaciones de Perusa (CNI-PG); (iii) una zona próxima al Instituto Experimental para la Olivicultura de Espoleto (lEO-ESP); (iv) una zona cerca Instituto Técnico Agrario de Pescia (ITA-PES); (v) una zona cerca de la Facultad de Agraria de la Universidad de Florencia (UNI-FL). De estas zonas han sido tomadas muestras de aceitunas de las variedades Frantoio, Leccino y Moraiolo, las cuales han sido procesadas con un micro-molino de aceite. Los resultados de la investigación, han puesto en evidencia que los parámetros de la calidad, tipismo y vida útil, y el flavor de los aceites estaban bastante influenciados por la zona de origen, es decir por los factores climáticos y pedológicos del medio agrícola. Se debe subrayar la influencia observada en algunos grupos de componentes, como fenoles, tocoferoles, compuestos volátiles aromáticos y ácidos grasos, puesto que estas substancias están estrechamente relacionadas con la calidad y el tipismo del producto. Algunos parámetros de la pureza están también influenciados por las variables ambientales.


INTRODUCTION
Climatic and pédologie factors, cultivation and agronomic techniques, harvesting, carriage and storage systems of olives, ripening degree of drupes, genetic factors (cultivar), and processing techniques, affect the analytical characteristics of oil (Bolinas and Angerosa, 1989;Olias, 1992;Bianchi and Pozzi, 1994).They also are influenced by the kind of container used to store the product (Ranalli, 1989).
The composition of oil is in large measure attributed to the agronomic or geographical origin (Servili et al., 1990;Osman et al., 1994).In fact, certain zones are renowned for the characteristics of their oils.In order to certify the origin zone of typical oils with excellent characteristics, Italy issued the law 169/92 (on the COD, controlled origin denomination) and European Economic Community (EEC) issued the Regulations 2081/92 (on the POD, protect origin denomination, and PGI, protect geographical indication) and 2082/92 (on the specificity attestations).These acronyms represent certification marks which are reported on the labels of containers.In this way virgin oil is valorised and its quality protected.
The relationships between analytical oil variables and pédologie and climatic variables have been less studied (Ferreiro and Aparicio, 1992;Pannelli et al., 1993).According to Ferreiro and Aparicio (1992) some sterols, triterpenic alcohols and hydrocarbons are correlated negatively with altitude.The influence of elevation variable on the composition of virgin oils was earlier pointed out by Armanino et al. (1989) and later by Osman et al., (1994); the latter authors observed that the oils from 100 m altitude were higher in phenols and unsaturated fatty acids, as well as had a higher oxidative stability and free acidity compared to 400 m elevation.The unsaturated fatty acid variable appeared also correlated with low temperatures of the cultivation location (Kiritsakis and Markakis, 1987).The positive influence of low temperatures on the oleic acid and unsaturated fatty acids/saturated fatty acids ratio values was also noted by Lotti et al., (1982) and Fiorino and Nizzi Griffi (1991).Pannelli et al., (1993) showed that rainfall was correlated negatively with the total oil phenol content and positively with total volatiles.Servili et al., (1990) found that the skeleton percentage of soil was correlated positively with phenol content of oils (possibly because it influences negatively the moisture percentage of soil).Other correlations were checked by Angerosa et al., (1996).
Notwithstanding some of the authors, using multivariate statistics, were even able to differentiate the oils produced under different levels of some agronomic variables, the scientific explanation of the aforementioned relationships has not given yet.
The effects of the environment/cultivar combination are very complex and play a basic role in determining the olive oil quality.The other factors can only modify the primary effects.These, and noticeably the relationships between the two above kinds of variables, have not yet been elucidated well and further studies are needed.
We are accomplishing a pluriannual investigation, whose first results, obtained in 1996, are given in this paper.On the other hand, several Italian virgin olive oils some of which are produced in well-known zones, have not yet been characterised well.Thus, it is not easy to get them a certification mark by the competent institution.This paper pretends to contribute to the solution of the problem.

Geographical location and olive growing features of the selected areas.
Three areas located in Umbría and two located in Tuscany, two central Italian regions where virgin olive oils of good quality are produced due to both good environmental characteristics and good processing technologies, were chosen.
The olive varieties considered in this study are cultivated in both the regions and this is the reason that led us to select them.Most of the Italian regions have autochthonous varieties.Frantoio, Leccino, Pendolino and Maurino olive varieties are the most cultivated in Tuscany, while Moraiolo and Dolce Agogía are the most cultivated in Umbria.The Spanish cv.Gordal is also present in some areas of the two regions.Leccino and Dolce Agogia are also employed as table olives.

Pédologie, climatic and cultivation parameters of the five origin areas.
The codes and the mean annual values of climatic parameters (concerning 1996) and those of pédologie ones are given in Table I.Regarding cultivation parameters, manuring, in the area of UNI-PG, is done using 2.7 Kg N, 0.5 Kg P2O5 and 2.4 Kg K2O per quintal olives produced every year.In the (c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://grasasyaceites.revistas.csic.esarea of ATI-PES, nitrogenous manuring is performed employing 2.0 Kg N per quintal olives produced every year, while phosphatie manuring is effected using (per Ha) 130 Kg P2O5 every three years, and potassic manuring is performed occasionally.In the area of NRC-PG, manuring is done employing 2.5 Kg N, 0.4 Kg P2O5 and 2.6 Kg K2O per quintal olives produced every year.In the area of UNI-FL, nitrogenous manuring is effected using 150 Kg N every year and 160 Kg P2O5 every two years, while potassic manuring is performed rarely.In the area of EOI-SPOL, manuring is done employing 2.7 Kg N, 0.3 Kg P2O5 and 2.3 Kg K2O.In all five cultivation areas, pruning is effected every 1-2 years and irrigation is not performed; 2-3 harrowings per year and antiparasitic treatments vs. olive fly, black scale and olive leaf spot are done; olive threes having a vase shape with 5x5 distances are cultivated; in 1996, in every area, the olive production was very low.Thus, in that year, in the five areas, the cultivation techniques used were very similar, while the climatic and pédologie conditions were dissimilar.

Maturation index, harvesting and processing of the olive samples.
In the five olive areas, 5 trees (approx.25 years old) with uniform characteristics were chosen for drupe sampling; a sample of 20 Kg was collected (4 Kg/tree) in each olive grove zone for each variety.The Moraiolo olive sample was not obtained from ATI-PES area.Harvesting was done by hand at different dates, during the maturation period, whell a same variety reached a similar ripening degree value in the different cultivation areas, as this parameter affects greatly the characteristics of oil.From each of the 20 Kg olive samples, two sub-samples, each weighing 1 Kg, were drawn.For each sub-sample, by applying the method proposed by the Estación Experimental de Olivicultura y Elaiotecnia de Mengibar, Jaén, Spain (Uceda étal., 1980;Hermoso étal., 1997), the ripening degree was determined.The pulp/stone ratio and the percentage of oil, moisture and solids were also assessed.The values averaged of olive maturation extent and of other drupe characteristics are given in Table II, together with the dates of olive sampling, and the codes of both samples and fruit features.
After harvesting, the olive samples were rapidly transported to the Istituto Sperimentale per I'Elaiotecnica, Pescara, were the drupes were subjected to the above measurements and processed.The fruits were wholesome, not damaged, and fresh.
To process the remaining 18 Kg olive samples to obtain oil samples for analyses, a mini oil-mill (Valpesana S.r.L, S. Casciano Val di Pesa, Italy), equipped with a hammer crusher, a mixer and a basket centrifuge was used.Kneading step lasted only five minutes, and was conducted without warming the paste.The rotary speed of the centrifuging drum was not higher than 2000 rpm.No usage of lukewarm water was done to fluidize the paste during the centrifugation step.

Oil and olive sample analyses
For the oil samples, by applying either analytical Community methods (E. C. Regulation N.° 2568/91) or non-official ones detailed in previous papers (Ranalli and Angerosa, 1996;Ranalli and Serraiocco, 1996 a,b;Ranalli and De Mattia, 1997;Ranalli et al., 1997), the following determinations were performed: -high-resolution gas-chromatographic (HRGC) analysis of the phenol fraction by using a 25 m X 0.32 mm i.d.capillary column coated with 0.1 jim of SE-54 (Lab Service Analítica Ltd., Anzola Emilia, Bo, Italy), after extraction with methyl alcohol.The resulting methanolic extract was concentrated to dryness by rotary evaporator, and the residue was subsequently recovered with 10 ml acetonitrile.This solution, after three washings with hexane, was evaporated in vacuum at a temperature below 35 "^C.The residue was dissolved into acetone, and 150 |Li l of bis(trimethyl) thfluoroacetamide were added to 1 ml of this solution.After 1h, the injection into the gas chromatograph was carried out.The internal standard was resorcinol (> 99% pure).Free tyrosol and hydroxytyrosol and tyrosol-and hydroxytyrosol-aglycones (dialdehydic forms of elenolic acid linked to tyrosol or hydroxytyrosol) have been identified in the HRGC chromatogram of phenolic fraction (Angerosa etal., 1995); -colorimetric determination of total phenols as well as of odiphenols, developing the colour with the Folin Ciocalteus's reagent in the first case and with the Arnow's reagent in the second case and taking the absorbance values at 725 and 450 nm, respectively; -high-performance liquid chromatographic (HPLC) analysis of tocopherols with a directphase column using a hexane-propan-2-ol (98.5:1.5, v/v) eluent and a UV detector at 292 nm wavelength.This fraction included essentially a-tocopherol, while p-, y-, and 5-tocopherol were present in minute amounts; -HRGC analysis of headspace volatile aromatic fraction, by using a 25 m x 0.32 mm i.d.capillary column coated with 0.20 |Li m of carbowax 20M (Nordion Ltd., Helsinki, Finland) after its extraction by stripping at 37 °C in a nitrogen stream and entrapping by activated charcoal.
The overall quality index (OQh) was evaluated by the algorithm developed by Olive Oil Council (0.0.1., 1990), based on the acid values, peroxide index and K270 (UV specific extinction at 270 nm wavelength), as well as the sensory score; the other overall quality index (OQb) involved also the oil polyphenol content (Solinasefa/., 1992).
To perform the quantitative descriptive sensory analyses, the COI test included in the E.C. Regulation N.° 2568/91 (B.C., 1991) was applied, by using a standard profile sheet.This is at present being modified by the COI.The score for each attribute was obtained by using a scale from 1 to 5 and was the result of the whole gustatory -olfactory -tactile perception.A fully trained analytical taste panel made up of 12 assessors, recognized by the CO!, and a sensory laboratory were used.The panelists had more than 8 years of experience in evaluating all types of olive oil (extravirgin, virgin, lampant, etc.).Oil samples were heated at 30 °C by a thermostat before analyses and were presented fully randomised to the tasters.The magnitude of several sensory attributes, such as olive fruity (ripe or green), apple, other ripe fruits, bitter, pungent, sweet, other allowable, sour/winey/vinegary/acid, rough, metallic, mustiness/ humidity, muddy sediment, fusty, rancid, and other unallowable attributes was assessed.Next, an overall evaluation of the magnitudes of positive and negative (off-flavors) attributes was made and, by means of a grading scale, the sensory score obtained.A structured scale ranging from 1 to 9 was adopted.
All oil sample determinations (as for drupe samples) were run in duplicate and the figures averaged.
The percentage of olive paste moisture was determined by oven method keeping at 105 °C a 50 g sample until constant weight.The residue was utilized for determination of the oil percentage, which was carried out by using Soxhiet apparatus and petroleum ether (b.p. 40-70 °) as the solvent.The paste solid percentage was evaluated by the formula: 100-(oil% + moisture%); the pulp/stone ratio of drupes was determined by using a laboratory blender for their depitting and washing the obtained stones with petroleum ether (b.p. 40-70 °).

Statistics
To process the data, after their normalization, multivariate statistiscs were applied, such as Principal Component Analysis (PCA) and Partial Least Squares regression (PLS2).The cross-validation procedure was used to determine the maximum number of significant dimensions to avoid data over-fitting.The multivariate evaluation of the effects of the experimental factors were made.The statistical package Unscrambler II Version 5. 52 /CAMO A/S, Trondheim, Norway) was used.For Hierarchical Cluster Analysis (HCA) the statistical Constat software was utilized.

RESULTS AND DISCUSSION
The PCA of the agronomic data set (concerning climatic and pédologie variables) (5 x 18 matrix) showed that the two first principal components (PCs) explained approx.70% of the total variance.The score-plot (Figure 1) by the dimensions 1 and 2 showed that the three Umbrian olive areas were well-differentiated, while the two Tuscany areas, which were positioned along the second component, were not very discriminated.Figure 2 shows the loadings for the dimensions 1 and 2, that point to which variables contributed to the discrimination of the olive areas.The PCA of the data set of characteristics of drupes (14x5 matrix) showed that the first two eigenvectors accounted for approx.90% of the total variance.The PCA scores of the objects on the 1 and 2 dimensions (Figure 3) indicated that the olive varieties, independently of cultivation areas, were clearly discriminated along the two first PCs (Frantoio and Moraiolo on the negative half and Leccino on the positive side).This was showed also by the Hierarchical Cluster Analysis.In fact, the dendrogram (not shown) displayed three great blocks each constituted by a same variety, with a similarity percentage within each block more than 65%.Each block was divided into sub-blocks depending on the cultivation areas, characterised by similarity percentages up to 95%.The loading-plot (not shown) from the first two eigenvectors indicated that the ripening index, oil %, and moisture % of drupes (Table II) were important in describing the first component, while the pulp/stone ratio and solids % variables were important in describing the second component.By comparing the score-plot with the loading-plot it was noticed that the ripening index, pulp/stone ratio and moisture % variables were mainly responsible for the discrimination of the Leccino variety, while the Frantoio variety was mainly differentiated by the solids % variable, and the Moraiolo variety by the oil % variable.
The PCA of the original analytical oil data led to identify three valid PCs which accounted for approx.54% of total variance.The 2D and 3D score-plots (not shown) by the first three PCs showed that both origin area and olive variety were not welldifferentiated.
By means of loadings, a screening of the original analytical oil variables was made, and those selected, accounting for a higher variance percentage, are reported in Table III, together with their values (averaged for variety), standard deviation, and codes.The PCA was applied to the reduced 14 x 20 matrix and the model improved as the first three significant components explained approx.80% of total variance.To explore the relationships between climatic and pédologie parameters and the N.° 20 analytical oil variables selected by PCA, the Partial Least Squares in latent variables (PLS2) was used.Three PCs was found to be statistically significant, which accounted for more than 70% of total Y-variance.The X-and Y-loadings for PC 1 vs. PC2 (Figure 4) indicated the following: (i) waxes (C40 + C42 + C44 + C46) could negatively correlate with organic matter of soil; (ii) fruitiness, oleic acid, linoleic acid, sensory score, 0QI1, 0QI2, and total sterols could negatively correlate with limestone percentage of soil; (iii) a-tocopherol, pleasant volatiles, oleic acid/linoleic acid ratio and unsaturated/polyunsaturated fatty acid ratio could indirectly negatively be influenced by windiness, while this variable could positively influence o-diphenols, hydroxytyrosol aglycones and total phenols (as it reduces the humidity of soil), in accord with Servili et al. (1990) and Pannelli et al. (1993); (IV) fruitiness could also negatively be affected by maximum air temperature and C/N ratio of soil.There were other correlations the above loadings enabled to hypothesize, but it is wise to wait for further data we will obtain in the next years before drafting conclusions.The relationships between the trend of climatic variables (especially rainfall and temperature), during growth and ripening of drupes, and analytical oil variables, we will also explore.The estimation of effects of the 14 objects (cultivation area/olive variety combinations) on the N.° 20 analytical oil variables was made.There was a fractional factorial design as the PESMo olive sample was missing.To validate the model, leverage correction was used, as there are not replicates.The main effects confounded with two-way interaction effects were estimated.By running PLS 1, the effects for each Y-variable were estimated.Factorial designs give only one PLS component per Y-variable (because all variables vary equally).This means that the systematic variation of the data set is completely described by one PC.For each Y-variable a normal distribution plot was obtained, showing the significant and the non-significant effects, the latter being those that were located along the straight line through (0;50).To interpret the results, the p-coefficients were also used.The values of the estimated effects are two times the (3-values.Subsequently, a diagnosis of the model for each Y-variable was made.A new model, including only the significant effects, was made.The plot of residuals showed that they were practically all on a straight line through (0:50), i. e., were normally distributed and thus the fit was adequate.Finally, the plot of residual vs. predicted values of the reduced model showed that the residuals were scattered randomly around the zero line, i. e., there were no systematic patterns in the residuals.This means that there was no unexplained systematic variation in the response, due to systematic variation in the insignificant design variables.Again the model seemed to be satisfactory.Figures 5, 6, and 7 show, as an example, some of the plots and the (3-coefficients only for the total phenol Y-variable.The significant effects for the N.° 20 analytical Y-variables have been summarized in a Table (not shown).There were, as expected, many significant effects due to the environment/olive variety combinations.However, both cultivation area and olive variety were not sufficiently discriminated, i. e., there was not a clearly overall predominant effect exerted by one of the two factors.However, the elevated standard deviation value recorded for some variables (e. g. total phenols, odiphenols, hydroxytyrosol-aglycones, pleasant volatiles), by averaging the analytical figures for variety (Table III), suggested that they were preponderantly affected by the cultivation area factor.

Figure 3
Figure 3Score-plot of cultivation area/variety combinations by the dimensions 1 and 2 from PCA of drupe variety features Figure 4 PLS2 loadings of selected analytical oil vanables (Y) and climatic and pédologie variables (X) for latent components 1 and 2 Figure 5 p-coefficients of cultivation area/olive variety combinations for oil phenol content variable of reduced model

Table Codes of the 20 selected analytical oil variables and their mean values ± SD for each olive variety
(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://grasasyaceites.revistas.csic.es