Viticulture is a highly profitable agricultural activity, nevertheless, winegrowers’ financial sustainability depends on good yield and on the high quality of the harvested products. These features are determined by several factors, among them, plant nutritional balance and soil fertilization. Brazilian viticulture applies fertilizers and soil acidity correctives that account for 38% of expenses in crop inputs and for approximately 16% of costs in the total maintenance of producing grapevines (>4 years) [1
]. Thus, there is high demand for methodologies to help winegrowers with soil corrective and fertilizer management, to achieve satisfactory grape, must, and wine yield, and quality [2
Soil analysis is globally used to estimate nutrient availability; the recorded values are mainly related to culture growth and yield parameters [3
]. However, it is essential carrying out leaf analysis for perennial fruit plants, such as grapevines, rather than soil analysis alone. This is because, compared with annual crops, trees overall record higher dry matter rates, accumulate more nutrients, and occupy a larger volume of soil [5
]. Accordingly, fruit trees acquire some nutritional stability at the adult phase [7
]. Owing to such a stability, the leaf composition diagnosis, allows adjustments of fertilization programs. In grapevines, leaf collection is suggested at full flowering or at berry veraison [8
The compositional nutrient diagnosis method (CND) takes into account the association between a given nutrient and the geometric means of concentration, on a dry matter basis, recorded for the other nutrients (multivariable relations), including those that are not analytically determined. It is considered as the best way to express balance in plant tissue [9
]. This technique was developed in Canada [9
] and has been applied worldwide to several annual cultures such as beans, maize, rice, soybeans, tomato, and Aloe vera
]. However, the methodology also has a potential high success in forest essences such as eucalyptus [16
], and in fruit plants such as banana, orange, pear, mango, and guava trees [18
]; but, it has not yet been applied to grapevines.
In 2018, Melo and collaborators [22
], used the diagnosis and recommendation integrated system (DRIS) to determine normal nutrient ranges in vineyards in Southern Brazil. However, the DRIS values were not associated with yield. Thus, it became mandatory to apply other methods, such as the CND, that are more accurate in estimating normal nutrient ranges in grapevines, in order to find good association with yield. The compositional nutrient diagnosis method generates a correlation factor for any nutrient by adding all essential elements in a multi-nutrient analysis. It also enables attributing the same weight unbalance to deficiencies and excesses; such a factor can be detected by applying the Mahalanobis distance [23
]. Besides, the CND methodology has only one standard deviation, and allows identifying and excluding atypical data (outliers), a fact that increases its reliability in result interpretation [18
]. Consequently, this method can be applied to a database set for a single region, to establish an association between grapevine nutritional status and yield.
We hypothesized that it would be possible to generate a range of specific sufficiency for the type of Vitis vinifera grape. The aim of the current study was to use CND standards, to determine the critical level and sufficiency band of nutrients in Vitis vinifera L. grapes.
3. Results and Discussion
The database comprised 81 commercial vineyards focused on grape crops for wine production, it held yield results and leaf nutrient concentrations that have presented yield variations ranging from 69 to 0.4 t ha−1, a mean of 14 t ha−1, and a standard deviation of 12 t ha−1.
The correlation matrix was initially used to explore the results in order to assess the correlation and the appropriate coefficient of determination between isolated nutrient concentration and yield. Accordingly, it was possible to observe the influence of a single-nutrient leaf concentration on the concentration of other nutrients and/or grapevine yield. The present study was not capable of assessing, through univariate analysis, any significant correlation, with high determination coefficient (Table 1
). Moreover, it is worth highlighting the low effectiveness of the concentration of a single nutrient on vine yield prediction [7
Remarkably significant correlations between nutrients and determination coefficient were observed between Ca and Mg, P and S, N and S, N and P, P and Mg (Table 1
). These correlations between nutrients indicate that changes in the concentration of a given nutrient in the plant can consequently change the concentration of others [18
]. This finding helps to explain why univariate correlations are not enough to justify the yield rates.
Reference population division was carried out based on Khiari et al. [35
], whose nutritional diagnosis for the 63 vineyards presented a yield mean inflection point, at cumulative function of 11,111 kg ha−1
. This value was applied to determine the reference high-yield subpopulation (N = 29) (Figure 1
Among the 63 vineyards, only 29 (46%) presented grape yield higher than 11,180 kg ha−1 (first production after the inflection point), and these composed the high-yield subpopulation. The other 34 vineyards (54%) composed the low-yield subpopulation.
The division into high and low yield populations allowed observing that none of these populations (Table 2
) exceeded the amount of correlation found in the complete database (Table 1
). This outcome was expected, since observation partitions generated a smaller number of occurrences. Moreover, there was no positive correlation between a single nutrient and yield, even in the high-yield population; this finding evidences that the appropriate balance among all nutrients ensures higher yield. The main associations highlighted in the complete database (N = 63) remained significant due to population division into high and low yield (Table 2
). They were also completed by the S-Cu association in the high-yield population and by the Ca-S association in the low-yield one.
Population division into high and low yield did not ensure the safe indication of appropriate nutritional content bands.
It is possible to point out that the minimum, maximum, and mean values of nutrients in leaves assessed in the high and low yield populations (Supplementary Materials available, Table S1
) did not show differences in their classification in the interpretation of appropriate nutrient concentration bands, based on Brunetto et al. [41
]. In addition to the correlations between nutrients and yield (Supplementary Materials available, Table S1
), the findings described above evidenced the need of using bi- or multivariate methods to diagnose nutritional status. It is so, because the average of nutrient concentrations in any of the assessed populations did not explain the yield rates recorded for the vineyards, but the ratio of each nutrient in regards to balance. It is also possible to highlight that all other nutrients had a normal interpretation, except for Mn and Zn, which showed excessive contents.
It was observed that the univariate indication does not properly represent the determination coefficients, not even the significant correlations (Table 2
). The indication of association between the CND-r2
and the Mahalanobis distance in the reference population can be observed in Figure 2
. This outcome evidenced that the longer the distance (D2), the greater the nutritional balance (CND-r2
). Similar results were reported in a study on potato plants [42
], where the reference population also showed a great nutritional imbalance (R2
= 0.34), showing that populations with adequate productivity may have the potential to improve further.
With regards to the present database, despite the high yields, there were nutritional disturbances, mostly because a significant part of observations were concentrated in the quadrant presenting the longest Mahalanobis distances (Figure 2
). Thus, it is important to notice that winegrowers must make soil and leaf analyses on a yearly basis and compare the results to CND-r2
standards, although it is also possible to observe some unbalance in the reference population.
The CND technique generated standards and statistical parameters of CND indices for grapevines (Table 3
). Thus, all samples can be compared to the standards, which comprise the mean and standard deviation of the high-yield population. This comparison generated indices to each nutrient.
It was possible to find IN, IP, IK, ICa, IMg, IS, IB, IFe, IMn, and IZn indices based on the mean concentration, recommended as appropriate for grapevines [41
], and by analyzing the mean normal concentration in leaf samples to set appropriate standards (Table 3
), assessed through CND standards. Results showed a CND-r2
= 63.72. Concentration of N, Fe, and Zn were underestimated under the herein tested conditions based on standards set for the CND method (Figure 3
); in other words, by following the recommendations one would find a shortage of these elements in comparison to the high-yield population. On the other hand, the concentrations of some elements, such as K, Ca, Mg, and S, were overestimated (Figure 3
); this outcome points out that the aforementioned recommendations may induce the excess of fertilization with these nutrients.
Similarly to the insertion of the mean concentration values recommended as appropriate to grapevines [41
], which relies on the approximation of nutrient concentrations found through univariate experimentation, winegrowers can use this instrument by inserting leaf analysis of their vineyards, and accomplish evaluations in compliance with reality in their property. This process would allow the application of actually necessary nutrients through fertilization, which would enable high yield, decrease the potential of soil and water contamination [18
], and ensure higher profitability, since it would allow the application of nutrients really demanded by grapevines.
There was highly significant correlation between the concentrations and indices of all nutrients (Table 4
) and between CND-r2
index and yield in all vineyards, as showed by Pearson’s correlation coefficient (R2
= 0.006), with decreasing adjustment. Thus, nutrients all together only explain a little of the recorded yield; this outcome goes against that recorded by Nowaki et al., [14
], who used CND in a tomato crop and found 0.17 adjustment. These contradictory outcomes likely happened because the assessed vineyards grew different cultivars [43
]. This finding suggests specific standards for each cultivar.
Critical nutrient levels in grapevine leaves were similar to those observed by Melo et al. [22
], (Table 5
). The establishment of sufficiency bands allowed comparisons to the literature about the herein assessed culture (Table 5
Compositional nutrient diagnosis standard levels were related to levels reported for grapevines grown in Rio Grande do Sul State [22
], where the present study was carried out. They were also related to outcomes found by Quaggio and Raij [44
], in São Paulo State, Brazil. There are great divergences about standards recommended in the handbook guiding the biggest grape producer state [41
], on the appropriate contents of N, Ca, Mn, and Zn. The greatest divergences were observed in the bands of Mn and Zn concentrations. These elements derive from fungicide applications carried out to control and prevent leaf and cluster diseases [45
]. However, they can also result from mineral fertilization, including contaminants, as well as organic fertilizers applied in vineyards as macro-nutrient source, mainly of N, P, and K [46
Overall, the CND method showed nutritional band amplitude closer to that observed by Melo et al. [22
]; however, it was smaller than that reported in the literature (Table 5
). Serra et al. [48
], considered the smallest amplitude band as positive information, because it allows greater accuracy to understand leaf content outcomes. This process reduces the possibility of dealing with low incomes, but it does not impair plant nutrition.