An exhaustive leaf sampling, according to the sampling intervals shown in Table 2, was carried out from pre-flowering to post-harvest. Thirty complete, diseases free and non-senescent leaves were sampled in each replicate. Sampling was carried out between 10 a.m. and 12 a.m., taking one leaf per plant, proceeding from fruit-bearing shoots of average vigour, out of 30 vines chosen at random within each replicate. The leaves collected were opposite to the first bunch from pre-flowering until pre-veraison, and opposite to the second bunch from veraison to ripening (Romero et al., 2010). Sunlight exposure over the canopy was also considered at sampling time and, therefore, leaves were sampled alternatively on both sides of the trellis. Samples were codified using the percentage of time that had elapsed in each phenological stage (Table 2).

Leaf blades and petioles were separated, washed three times with tap water, and then rinsed with distilled water and oven-dried (Dry-big; J.P. Selecta, Barcelona, Spain) at 70ºC for 48 hours. Finally, dried samples were ground and homogenized with an ultracentrifugal mill (ZM1; Retsch, Haan, Germany), in order to pass through a 0.5 mm mesh.

For total N determination (N-organic+N-NH₄ ⁺) in leaf blades and petioles, 0.20 g of the ground sample was used to carry out a wet digestion using the Kjeldahl method (Jones et al., 1991). For the remaining nutrients (P, K, Ca, Mg, Fe, Mn, Zn, Cu and B), 0.20 g was used to carry out a wet digestion with 95% H₂SO₄ and 30% H₂O₂ (Hoenig et al., 1998) and afterwards were analyzed by inductively coupled plasma-optical emission spectrometry (Optima 3000DV, Perkin Elmer, Norwalk, CT, USA). Double deionised water (Milli-Q; Milli-pore, Bedford, MA, USA) was used for all dilutions. Concentrations were expressed on a dry weight (DW) basis.

A principal components analysis (PCA) was performed to detect the vectors that differ from the ten dependent variables (N, P, K, Ca, Mg, Fe, Mn, Zn, Cu, B) data. A non-rotated component matrix was used as factor analysis and the extraction method was performed by a scree plot with eigenvalue>1. The variables were adequately represented within the model. Furthermore, comparisons for equality of means, between blade and petiole tissues at each sampling time, were performed using the parametric t-Student test.

Nutrient trends were fitted with Tablecurve 2D (vers. 5.1; SYSTAT Software Inc., San Jose, CA, USA). The coefficient of variation (CV) percentage between replicates was analyzed for each nutrient, tissue, and sampling date as a measurement of the reproducibility of the analysis within each vineyard.

Finally, the differences between consecutive sampling dates were determined by means of a monofactorial ANOVA followed by a post-hoc test (Tukey, p≤0.05) in order to define the periods of validity for references at different phenological stages. The ANOVA procedure was performed by comparing groups of five consecutive samplings, as a moving window throughout the growing season, displacing the ANOVA procedure throughout the period to the next sampling and discarding the first sampling of the previous ANOVA to assemble the new group of five samplings to perform the next ANOVA procedure. Furthermore, when a long period with no significance differences was detected, the first sampling of the first window was additionally compared with the last four samplings of the last window to confirm the absence of differences over the whole period. This procedure prevents error accumulation as it increases the number of comparisons within the ANOVA.

Statistical procedures were performed with SPSS (vers. 15.0; SPSS Inc., Chicago, IL, USA) and Statgraphics Plus (vers. 5.1; Manugistics, Inc., Rockville, MD, USA).

PCAs are employed to reduce a large set of inter-correlated variables to a smaller number of uncorrelated variables. When applying this method, if the variables are highly correlated, some of the first components are able to describe most of the data variability. In our study three principal components (PC) contained approximately 83% of the total variance of the ten original variables (Table 3). The first PC (PC1) explained 41% of the total variance for both vineyards. The second component (PC2) explained 30% of the variance for the RA and 32% for the RM dataset. The third component (PC3) explained 12.6% of the variance for RA and 10% for RM.

The highest component loadings for the RA vineyard were: P, K, Ca, Mn and B for PC1, N, Mg, and Fe for PC2, and Zn for PC3. On the other hand, the highest component loadings for the RM vineyard were: K, Mg, Fe, Zn, and Cu for PC1, N, P, Ca, and Mn for PC2, and B for PC3. The PCA for the RA dataset showed the existence of a strong relationship between PC1 and the phenological stage as well as between PC2 and the analyzed tissue (Fig. 1A). Inversely, for the RM dataset, PC1 showed a strong relationship with the tissue and PC2 with the phenological stage (Fig. 1B). These two components together explained more than 71% of the total variance in both cases.

The positive correlations between the original nutrient concentrations and PC2 for RA, and between those concentrations and PC1 for RM, indicated that nutrient concentration in blade was higher than in petiole (Table 3). Therefore, blades showed higher concentrations than petioles for N, Fe and Cu, and slightly higher for Ca and B (Figs. 2 & 3; Table 4). Consequently, petioles had higher concentrations of K, Mg, and, to a lesser extent, Mn, and Zn. Phosphorus did not show clear behaviour as it had higher concentration in blades from RA but lower in blades from RM (Table 4; Figs. 2C,D).

Concerning the phenological stage, component loadings of the original nutrient concentrations for PC1 for RA, and for PC2 for RM (Table 3) indicated that the nutrients which positively correlated with the PC showed an accumulating trend in the analyzed tissue throughout the growing cycle (Table 3; Figs. 1C,D). That was the case for Ca, Mg, Fe, Mn, and Cu; while N, P, K, and B showed a negative correlation with the PC and, therefore, a reduction in the concentration in both tissues throughout the growing season. Finally, Zn showed a low correlation with respect to the PC, which indicates that the Zn requirements were relatively unsteady throughout the crop season in both tissues, mainly for the RA vineyard (Table 3; Figs. 3E,F).

The third principal component (PC3) explained 12% and 10% of the global variance, showing a low influence in the total variability compared to the other PCs, phenological stage and tissue (Table 3).

Nitrogen showed maximum concentration at pre-bloom sampling. Its concentration decreased over the course of the growth cycle until a stabilization trend for blade was noted around the veraison phenological stage (Figs. 2A,B). Phosphorus also showed a declining pattern of concentration from flowering to harvest (Figs. 2C,D).

Potassium showed a declining pattern for blades (Figs. 2E,F). However, petioles only showed this pattern for RA vineyard (Fig. 2E), while petioles for RM showed an irregular trend throughout the growth cycle (Figs. 2F). On the other hand, Ca and Mg showed an increasing concentration in both tissues (Figs. 2G-J). In general, macronutrients showed a nutrient evolution adjusted to a logarithmic curve in blade and petiole, with a decreasing slope until values near to zero when the veraison phenological stage was reached (Figs. 2A-J).

With respect to micronutrients, trends were not as easily recognizable as for macronutrients (Fig. 3). Some micronutrients showed a logarithmic trend (B, Fe and Zn in petiole for RM); others showed a polynomial fitting (Fe and Cu in petiole and Zn in blade for RA, as well as Cu for RM); and the others fitted to a sigmoidal curve (Fe, Mn, and B in blade and Zn in petiole for RA). Those nutrients with a sigmoidal curve showed two periods where the slope tends to zero, the first around flowering and the second at veraison (Fig. 3).

Iron concentration increased throughout the cycle except at the beginning of bloom and at veraison for both tissues (Figs. 3A,B). Manganese concentration increased in both tissues until veraison, and then its concentration remained stable until the end of the growth cycle (Figs. 3C,D).

Zinc concentration in RA decreased until half of the berry growth phenological stage, increasing thereafter until the end of veraison and then maintaining concentration levels up to harvest in leaf blade and petiole (Fig. 3E). However, Zn concentration in blades from the RM vineyard maintained its levels after fruit set until the end of veraison, while the petiole showed a constant Zn concentration increasing from half of the berry growth until the end of veraison which denoted a different behaviour (Fig. 3F).

Copper maintained relatively constant levels of concentration in both leaf blades and petioles until flowering and even to early pea-sized berries. However it was noted that the application of phytosanitary sprays with a copper-based formulation prompted an increase in blade Cu concentration in the early maturation stage (Figs. 3G,H).

Finally, B concentration increased at the beginning of the cycle, afterwards showing a sharp decrease in blades (Figs. 3I,J).

A stability period is defined as the period where the nutrient concentration remains stable, allowing its comparison with a reference which, therefore, would be reliable for that complete period. However, the length of each stability period is difficult to be delimited only since the slope. Due to it, an stability period could also be defined by the absence of statistical differences between consecutive samplings.

Different stability periods of various lengths in time, were obtained for each nutrient and tissue because of the different behaviour of the nutrient concentration throughout the season. From these different stability periods, the period of validity for references proposed for flowering or veraison stages was determined. Furthermore, a period of sampling for blade or petiole could be also delimited for a common diagnosis of all the analyzed nutrients.

Fig. 4 shows the stability periods for macronutrients and micronutrients concentration in blades for both vineyards. Only those periods which included the flowering or veraison phenological stages are shown.

Blade results showed that there was no period which included the complete flowering phenological stage for any nutrient in the RA vineyard (Fig. 4) because of the absence of sufficient samplings within this phenological stage for this vineyard (Table 2). Nevertheless, for N, P, K, Mn, Zn and B a period which included the complete flowering stage was found for the RM vineyard. This period was limited to flowering for N and P, and was slightly broader for K and B. For Mn, this period was extended to K75, and even until complete veraison for Zn due to the high variability of the Zn concentration in blade at some moments during the growing cycle (Fig. 4). This latter fact was the origin of the absence of significant differences between samplings for Zn at RM.

The RM vineyard showed a period of stability which included the whole of veraison (the last sampling time of this vineyard) for N, P, K, Mn, Zn and B. However, the stability period for Ca ranged from 25% to 100% of veraison, and for Mg and Fe it comprised 50% to 100% of veraison (Fig. 4).

On the other hand, in the RA vineyard only Fe and Mn had a stability period that lasted all veraison while the other nutrients showed stability periods which ranged, in general, between M50 (50% of veraison) and N50, or 50% of ripening (Fig. 4).

The stability periods for petioles at both vineyards are shown in Fig. 5. For petioles at flowering, the nutrient concentration and thus the stability periods behaved similarly to blades. Petiole results showed that there was not a period which included the complete flowering phenological stage for N, P, Ca, Fe and B in the RA vineyard (Fig. 5). However, the absence of sampling for H100 and I40 from this vineyard (Table 2) rules out an adequate evaluation of the flowering stability period length for the RA vineyard. On the other hand, Ca, Mg and Fe had stability periods that did not include the complete flowering stage in the RM vineyard (Fig. 5). Therefore, Ca, Mg and Fe would need one reference at the beginning of flowering and another at the end due to the rapid change in the concentration of these elements in petioles (Fig. 5).

With respect to veraison, the RA vineyard showed a stability period which included this phenological stage completely for all nutrients except Ca, which ranged from M15 to 50% of ripening (N50). Nevertheless, the RM vineyard only showed a stability period for the complete veraison phenological stage for N, P and Fe. Furthermore in the RM vineyard, Zn and B showed a stability period from M15 until the end of veraison (the last sampling time of this vineyard), K from M50 to the end of veraison, and finally Ca, Mg and Mn only showed a stability period within M15 to M75 phenological stages (Fig. 5). In these cases, a reference obtained at M50 for Ca, Mg and Mn would not be valid for a sampling at the beginning (M5) or at the end (M100) of this phenological stage due to the change in the concentration of these elements in petioles (Fig. 5).

The CV is used to determine the most appropriate tissue for nutritional diagnosis. The maximum CV values for macronutrients were, in general, greater in petioles than blades. The greatest differences between both tissues were found mainly at veraison especially for N, K and Ca in the RM (Figs. 6B,F,H). At flowering, differences between both tissues were often lower than at veraison (Fig. 6).

Micronutrients did not show clear differences between blades and petioles. The maximum CV values were usually lower than 50% and, in general, higher in petioles than blades (Fig. 7). Furthermore, differences were found between vineyards. Iron, Mn and Zn showed greater CV in petioles from RA vineyard throughout the crop season (Figs. 7A,C,E) whereas results from RM did not show such marked differences.

The differences between both tissues for N has been described for varieties such as ‘Tempranillo’ (Romero et al., 2013), ‘red Grenache’ (Benito et al., 2013) and others (Christensen, 1984; Champagnol, 1990; Kliewer, 1991). Besides this, the higher concentrations in blades for Ca, Fe, B and Cu have also been observed in previous reports (Christensen, 1984; Benito et al., 2013; Romero et al., 2013). This is also the case for the higher concentration in petioles described in the bibliography for K (Christensen, 1984; Schreiner & Scagel, 2006), Mg (Schreiner & Scagel, 2006), Mn (Romero et al., 2010) and Zn (Christensen, 1984; Romero et al., 2013). However, the differences between both tissues were not so marked for K, Mn and Zn in the RA vineyard or for B in the RM plot. Finally, Atalay (1978) found a higher P concentration in petiole from vineyards with high P concentration in the soil, while Benito et al. (2013) observed that P concentration was higher in blades than in petioles. These results show the importance of proposing specific reference levels for blade and petiole tissues, separately.

The variation of the nutrient content in blade tissues throughout the crop season has been widely discussed by different authors (Colugnati et al., 1992; Parejo et al., 1992; Benito et al., 2013; Romero et al., 2013). In a PCA when one of the PC is highly related to the phenological stage, the original variables (nutrients) with sharp changes in concentration over the growing season are related to that PC.

Therefore, in our study, those elements showing slight concentration changes over the growing season present a weak relation with PC1 for RA and with PC2 for RM (Table 3). This result shows the importance of proposing specific reference levels for different phenological stages, mainly for those nutrients with strong evolutionary behaviour over the crop season.

In addition, our results showed that the vineyard origin and/or the year effect could modify the nutrient concentration. Other studies showed that the year effect was low on the total variance of datasets obtained for single vineyards throughout different years (Benito et al., 2013; Romero et al., 2013).

Most of the trends, mainly for macronutrients, showed curves with slopes near to zero at veraison (Fig. 2). Furthermore, some nutrients with sigmoidal behaviour curves presented two periods with slopes near to zero, one at the flowering phenological stage, which was shorter than that found at veraison (Fig. 3).

In general, seasonal trends for macronutrients in blades and petioles (Fig. 2) concurred with results found for other cultivars, such as ‘Chardonnay’ (Parejo et al., 1992), ‘Pinot noir’ (Schreiner, 2005), ‘red Grenache’ (Benito et al., 2013) or ‘Tempranillo’ (Romero et al., 2013).

The N and P behaviour is in agreement with results found by other authors (Christensen, 1984; García–Escudero et al., 2002; Schreiner & Scagel, 2006; Benito et al., 2013; Romero et al., 2013). The decreasing pattern for K was also found for Colugnati et al. (1992) and Porro et al. (1995). Furthermore, the increasing concentration of Ca and Mg in both tissues has also been shown by different authors (Stevens, 2005; Schreiner & Scagel, 2006; Benito et al., 2013; Romero et al., 2013).

Regarding micronutrients, the increases in Fe and Mn, agree with previous reports for other cultivars (Schreiner, 2005; Benito et al., 2013; Romero et al., 2013). The different trends observed by the Zn were probably due to either different climatic conditions and/or the soil effect (Figs. 3F,G). The trend observed by the Zn in RA was observed by Christensen (1984). Finally, the B pattern was also reported by Failla et al. (1995).

With respect to Cu, Schreiner (2005) observed a decreasing trend for Cu concentration in blade throughout the season for ‘Pinot noir’ cultivar while, in this study, blade increased Cu concentration in the early maturation stage. Therefore, the application of fungicides impeded the study of Cu behaviour (Benito et al., 2013; Romero et al., 2013).

Finally, the differences between nutrient concentration for leaf blades and petioles, as well as the seasonal changes throughout the growing season, confirm that it is necessary to design specific reference levels for each tissue and phenological stage.

The most adequate moment for nutritional diagnosis would correspond to the period in which nutrient concentrations remain highly stable for some time in the tissue analyzed allowing a reliable comparison with their references. Therefore, the best sampling times are those where the concentration trend shows a slope as near as possible to zero. Most of the trends, mainly for macronutrients, showed curves with slopes near to zero at veraison (Fig. 2), making that phenological stage an adequate moment to propose useful references. Furthermore, those nutrients with sigmoidal behaviour trends presented a shorter stability period at flowering (Figs. 2 & 3).

Flowering and veraison are the preferred phenological stages to evaluate the grapevine nutritional status (Robinson, 2005). Therefore, the available bibliography offers different reference values for nutritional diagnosis in grapevine tissues at these two phenological stages. However, the length of each stability period is not fully defined and it is difficult to delimit it considering only the slope, especially at flowering. Therefore, a statistical determination of the differences between sampling times becomes necessary to enclose the validity period for each nutrient reference at flowering and veraison.

Considering all constraints, the best time to perform a multi-element analysis of leaf blades and petioles during flowering for the ‘Graciano’ variety is at 100% flowering as it is an easy recognizable phenological stage, and references provided for this moment remain valid until fruit set. This recommendation agrees with results obtained for other varieties (Benito et al., 2013; Romero et al., 2013).

Regarding veraison, the recommendable period to carry out a general nutritional diagnosis of the ten nutrients studied for blades extends from 50% veraison to beginning of ripening, except for K. This situation is not similar to recommendations obtained for other varieties in longer monitoring studies, where the stability period covers all the veraison period or even allows an early diagnosis, up to a month before the start of this phenological stage (Benito et al., 2013; Romero et al., 2013). Therefore, the results suggest the need for more years of study to avoid the year and/or the soil effect on the nutrient concentrations in blades.

With respect to petioles, the optimum period for a common diagnosis at veraison ranged from two weeks after closed bunch until one week before the end of the veraison (75% of the berries coloured), except for K. Benito et al. (2013) and Romero et al. (2013), for ‘red Grenache’ and ‘Tempranillo’ varieties respectively, showed that the diagnosis in veraison with respect to a reference for 50% of coloured berries (M50), ranged from two weeks before the veraison to the beginning of ripening. This difference might be due to variety behaviour. However it is also necessary to extend the years of study to adjust these results for petiole. Finally, phytosanitary treatments did not permit a correct copper diagnosis throughout the growing season.

To evaluate the general nutritional status of a vineyard, the vine grower usually carries out a single analysis that compares the results to the standards. Therefore, an adequate tissue for nutritional diagnosis should provide acceptable analysis reproducibility (Wolpert & Anderson, 2007). A high variability between replicates from a uniform vineyard implies that obtaining representative nutrient concentrations is difficult. Due to that, the CV is used for the evaluation of the analysis reproducibility in both tissues at each phenological stage throughout the growing season.

Considering the diagnosis of individual nutrients, results suggest that blade was a more appropriate tissue than petiole to compare N, P and K concentrations with respect to their references (Figs. 6A-F). Calcium and Mg did not show clear differences between blades and petioles throughout the growing season, and CV in both cases was lower than 20%, but results suggest that blade would be preferable at flowering and veraison phenological stages (Figs. 6G-J). Some of these results agreed with those found for ‘Tempranillo’ (Romero et al., 2013) as for N and K at both phenological stages and for P and Mg at veraison. Furthermore, results are also in agreement with those found for N, P and K for ‘red Grenache’ cultivar (Benito et al., 2013).

The greater CV values in petioles implied a lower reproducibility of the analysis that may lead to diagnosis errors when the vinegrower compares a routine single analysis with the macronutrient reference values (Benito et al., 2013). Furthermore, CV result also suggest that the analysis of blade and petiole separately would be more appropriate than whole leaf since one tissue will always show a lower CV.

With regard to the micronutrients, results found for both vineyards are contradictory and therefore it is not possible to establish a blade vs. petiole preference sampling policy (Fig. 7).

In summary, the higher variability within petiole nutrient concentrations for the cv. Graciano suggests that, for routine single analysis, leaf blades would be preferable to petioles for N, P, K, Ca and Mg diagnosis at both flowering and veraison. With regard to micronutrients, the differences found between both vineyards do not allow to define the best tissue to analyze micronutrients. In general, references for blade and petiole at full flowering could be valid for diagnosis of all nutrients even in late samplings at fruit set. On the other hand, references obtained at 50% veraison, would allow the simultaneous diagnosis for leaf blades of all the studied nutrients between the middle of veraison and ripening. For petioles, the simultaneous diagnosis period ranges from the beginning of veraison until one week before the end of this phenological stage.

The methodology employed is appropriate in establishing the most adequate tissue and moment for the diagnosis of ten different nutrients; however, it is necessary to extend this study for a longer period to include data from different pedoclimatic subzones, in order to obtain conclusive results for ‘Graciano’ and to evaluate the local effects on nutrient behaviour.