INTRODUCTION
Chlorophylls are green pigments specialized for the absorption of sunlight, mostly in the blue and red regions of the visible spectrum used by plants to convert into chemical energy. In addition to chlorophyll a, other pigments are involved in photosynthesis, including chlorophyll b, carotenoids, and phycobilins, the latter three known as accessory pigments. Although chlorophylls are the most abundant pigments in plants, they undergo degradation as a result of environmental factors (light, solar radiation, and air temperature) and intrinsic factors such as metabolization during leaf senescence (^{Streit et al., 2005}).
Chlorophyll content gives a measure of leaf nitrogen because there is a close association between the pigment and the nutrient. Moreover, considering that chlorophyll concentration varies with the stages of plant development, its estimation in the field in different phenological stages of a crop may give the farmer information about the nutritional status of the plants that allows them to make decisions related to the monitoring and management of agricultural activities (^{Costa et al., 2015}; ^{Elarab et al., 2015}, ^{Arantes et al., 2016}). In addition, the monitoring of chlorophyll may be used as an indication of the plants' water status (^{Silva et al., 2014}). Thus, considering the importance of photosynthetic pigments in cultivated species and its relationship with the remaining production factors, the measurement of its content may be considered as relevant information to support crop management.
The quantification of chlorophyll content is usually performed using destructive methods such as UV and visible spectroscopy and high performance liquid chromatography, which require the conversion of leaf extract absorbance to values of pigment concentration through the use of standard published equations (^{Richardson et al., 2002}; ^{Prado-Cabrero et al., 2016}). These are time-consuming, laborious methods that depend on sturdy equipment; in addition, they destroy the available leaf samples, thereby precluding consecutive evaluations of the same plant material (^{Dey et al., 2016}).
However, there is the alternative of non-destructive optical methods. They are based on the use of hand-held meters that give immediate readings of the intensity of the green color of the leaves that are related to relative leaf chlorophyll content. This method allows repeated evaluations in the same site, to perform a larger number of measurements, and it simplifies the study of pigment dynamics throughout the crop cycle. Moreover, intensive monitoring using these instruments enables the characterization of spatial variability of chlorophyll content in cultivated areas.
The use of instruments and sensors that allow the intensive and continuous monitoring of variables of interest associated with spatial analysis techniques allow the adoption of precision agriculture (PA). In this context, the use of hand-held meters to determine leaf chlorophyll content and its relationship with leaf nitrogen content was shown by ^{Taskos et al. (2015)}. This demonstrates the applicability of these meters in the assessment of the variability of plant nutritional status and helps to make decisions regarding the potential implementation of variable rate management, a concept that is a part of PA.
The objective of the present study was to fit regression models to describe the temporal variability of the relative index of chlorophyll a (Chl_{a}), chlorophyll b (Chl_{b}), and total chlorophyll (Chl_{t}) obtained with the use of a portable meter, and to conduct the assessment of the temporal stability and mapping of Chl_{t} index in a commercial Chardonnay vineyard, in the period between the onset of flowering and full fruit maturation.
MATERIAL AND METHODS
The study was conducted in a commercial wine grape orchard (Vitis vinifera L.), cv. ‘Chardonnay’, with a total area of 1.10 ha, located in the municipality of Espírito Santo do Pinhal, state of São Paulo, Brazil (coordinates: 22° 10′ 49.1” S and 46° 44′ 28.4” W; mean altitude: 875 m). The total area of the vineyard under study was divided into two subareas, namely Area 1 and Area 2, which covered 0.60 and 0.50 ha, respectively. The region's climate was classified as B3rB'3a' (^{Rolim et al., 2007}), which was characterized as mesothermal, with absent or reduced water deficit and a sum of reference evapotranspiration varying between 855.0 and 997.0 mm per year. The soil in the experimental area was classified as eutrophic Tb Haplic Cambisol (^{Santos et al., 2018}).
The spacing adopted in both sub-areas of the vineyard was 2.50 m between rows and 1.00 m between plants. The grapevines were grafted on Paulsen 1103 rootstock, grown in a vertical espalier system, and trained using the unilateral Royat cordon system. Moreover, the grapevines were subjected to double pruning management (^{Dias et al., 2017}) with the aim of taking the crop into autumn and winter seasons when the climate conditions are more favorable for fruit maturation. Vineyard irrigation was performed using a drip system, with one lateral drip line per plant row and two emitters per plant, placed 0.50 m apart and with an average flow of 1.88 L h^{-1}. The criterion adopted for the irrigation management was the replacement of crop evapotranspiration (ETc, mm day^{-1}).
The measurement of the relative index of chlorophyll a (Chl_{a}), chlorophyll b (Chl_{b}), and chlorophyll a+b or total (Chl_{t}) were made in situ in leaves of georeferenced plants using a ClorofiLOG portable meter (model CFL 1030, Falker Automação Agrícola, Porto Alegre, RS, Brazil). This instrument emits light at three wavelengths, 635 nm and 660 nm in the red region and 880 nm in the near infrared region (^{Barbieri Júnior et al., 2012}). The emitted radiation is transmitted by the leaf and reaches a receiver that gives measurements proportional to chlorophyll absorbance in a dimensionless unit. Georeferencing of the plants from which the leaves were selected to measure the indices of chlorophyll were collected using HiPer GGD (TOPCON, Pleasanton, CA, USA) GNSS (Global Navigation Satellite System) receivers with RTK (Real Time Kinematic) centimeter correction.
A systematic sampling system was used to perform measurements at regular intervals of three plants along the planted rows, in interspersed rows, resulting in a grid composed of 346 and 303 sampling points in Area 1 and Area 2, respectively. The limits of both areas and the points in the sampling grids used for the in situ measurement of the Chl_{a}, Chl_{b}, and Chl_{t} indices in Chardonnay grapevine leaves using the described portable meter are shown in Figure 1.
The assessments were made during the 2017 grapevine cycle, which was initiated with the pruning on January 4, 2017 and ended with the harvest on May 16, 2017, thus lasting for a total of 132 days. The measurements of the chlorophyll indices were performed on seven occasions during the 2017 cycle (at 41, 57, 64, 78, 85, 99, and 120 days after pruning - DAP). The onset of flowering and of fruit maturation occurred at 37 and 92 DAP, respectively.
The chlorophyll indices were measured in two leaves per plant, positioned opposite the bunches (in the assessments conducted up to defoliation in the region of the bunches, at 76 DAP, for greater exposure of the bunches to the sun) and in the middle third of the grapevine canopy (in the assessments starting from 76 DAP). The criterion for the selection of the plant material to be analyzed was fully expanded leaves with healthy appearance.
Data analysis initially consisted of the application of descriptive statistics to the sets of values obtained for Chl_{a}, Chl_{b}, and Chl_{t} indices for each assessment date, which resulted in the estimation of measures of central tendency, dispersion, and distribution pattern. Still in the exploratory analysis phase, the dispersion of data points around the mean was classified based on the coefficient of variation (CV), which was classified as low (CV ≤ 15%), moderate (15 < CV ≤ 35%), or high (CV > 35%) according to the categories proposed by ^{Wilding (1985)}.
The assessment of the temporal variability of Chl_{a}, Chl_{b}, and Chl_{t} indices was performed by fitting regression models to predict and describe the variation trend of the mean values of the indices over time. The models were fitted according to the pattern of experimental data layout in a scatterplot, and the coefficients of these functions were estimated using the least squares method.
The temporal stability of the Chl_{t} index, measured in the various sampling points in both vineyards, was assessed through analysis of correlation between the values obtained in the same sites on different dates, according to Spearman's rank coefficient (r_{s}), calculated using [eq. (1)]:
Where:
n is the number of sampling points of the Chl_{t} value in each area;
X_{i,j} and X_{i,j′} are the Chl_{t} values measured in sampling point i on dates j and j′, in this order,
R(X_{i,j}) and R(X_{i,j′}) are the ranks of X_{i,j} and X_{i,j′} among all the values obtained on dates j and j′, respectively. The value of this coefficient varies from 1 to −1, which represent a perfect positive and negative association between ranks. Values of r_{s} close to 0 and 1 were indicative of the greater and lower magnitude, respectively, of the temporal stability of the Chl_{t} index between the two assessment dates. The descriptive statistical analysis and the regression analysis of the Chl_{a}, Chl_{b}, and Chl_{t} data, as well as the assessment of the temporal stability of the Chl_{t} index based on the r_{s}, were performed using the R software version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria, 2018).
In addition, geostatistical analysis was applied only to the Chl_{t} data set for each assessment date to evaluate the spatial variability of chlorophyll content by estimating empirical semivariograms and fitting theoretical models to the latter. Spherical, exponential, and Gaussian theoretical models were tested to fit the semivariograms. The criterion used to select the best theoretical model to characterize the spatial dependence of the Chl_{t} data was the value of the root-mean-squared error (RMSE) that resulted from the cross validation. Thus, the selected theoretical models were those with a lower RMSE. In addition, the modeling of the semivariograms showed the existence and correction of potential anisotropic effects associated with the spatial variability of the Chl_{t} index, as per the method proposed by ^{Oliver & Webster (2014)}. Once the models of the theoretical semivariograms were defined, the degree of spatial dependence was determined based on the spatial dependence index (SDI) proposed by ^{Seidel & Oliveira (2014)}, calculated independently for each model. Considering the spherical, exponential, or Gaussian models, respectively, the spatial structure of Chl_{t} index variation represented by spatial dependence was classified as strong (SDI > 15%, SDI > 13%, and SDI > 20%), moderate (7% < SDI < 15%, 6% < SDI ≤ 13%, and 9% < SDI ≤ 20%), or weak (SDI < 7%, SDI < 6%, and SDI < 9%) (^{Seidel & Oliveira, 2016}).
After the fitting of the theoretical models, the ordinary kriging method of geostatistical estimation was used for spatial prediction and generation of thematic maps for Chl_{t} index (^{Oliver & Webster, 2014}). The estimation of the Chl_{t} index in non-measured sites in both areas and in each assessment date by kriging allowed modeling of the distribution and spatial variability of Chl_{t} index across the phenological stages identified in the assessed period of the grapevine cycle. Subsequently, the estimated Chl_{t} data were classified into three distinct categories or regionalization, including high values (Class 1), intermediate values (Class 2), and low values (Class 3) of total chlorophyll content, assuming an ordinal scale as measurement of the variable in question. This categorization was made considering the natural breaks method, or Jenks optimization method, as the rule for the classification of the interpolation results. The geostatistical analysis described in this study was performed using the GS+ software version 7.0 (Gamma Design Software, Michigan, USA, 2004), and the thematic maps were created using the QGIS software version 2.18.17 (QGIS Development Team, 2016).
Finally, the comparison between the categorical maps of the Chl_{t} index obtained on subsequent dates was made based on the quantification of the level of spatial association between the regionalization of this variable, and the V-measure (V_{1}), a cluster evaluation measure, was used according to ^{Nowosad & Stepinski (2018)} (adapted). The values of this global measure of association vary between 0 and 1 and indicate a higher or lower spatial agreement, respectively, between the categories represented in the maps. In addition, the magnitude of the overlap between the regionalization of two consecutive Chl_{t} index maps was estimated using the Mapcurves algorithm proposed by ^{Hargrove et al. (2006)}, based on the calculation of the goodness of fit (GOF) for measuring the coincidence between its categories. The highest possible score for GOF is 1, which indicates an exact coincidence between the compared maps. Both mentioned analyses were performed using the R software version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria, 2018).
RESULTS AND DISCUSSION
The normality of the distribution of the chlorophyll index data on all assessment dates was tested using the similarity between measures of central tendency (mean and median), in addition to coefficient of skewness (Cs) close to zero. The variability of the data obtained on all assessment dates and in both areas was classified according to the values of CV (low when CV ≤ 15% or moderate when 15 < CV ≤ 35%).
Given the CV value (21.97%), the lowest homogeneity of the data distribution in Area 1 was related to the measurements of the Chl_{B} made at 41 DAP, i.e. at the onset of flowering, although variability was classified as moderate. In this area, the lowest CV (5.73%) was related to the measurement of the Chl_{a} index performed at 120 DAP at the end of the assessment period near full fruit maturation, 12 days before the harvest. In Area 2, the set of data with the highest dispersion was also related to the Chl_{b} value at 41 DAP, which was associated with a CV of 20.70% and resulted in moderate variability. In this area, the lowest relative variation of the data sets was similar to that obtained in Area 1 regarding the index with the lowest variability, because it was also related to the obtained Chl_{a} index but at 85 DAP before the onset of maturation, assuming a CV value of 4.31%.
DAP | index | n | Mean | Med | Min | Max | s | CV (%) | Ck | Cs |
---|---|---|---|---|---|---|---|---|---|---|
Area 1 | ||||||||||
Chl_{a} | 346 | 22.88 | 22.68 | 12.55 | 32.05 | 3.37 | 14.71 | -0.216 | -0.056 | |
41 | Chl_{b} | 346 | 5.96 | 5.95 | 2.95 | 9.75 | 1.31 | 21.97 | -0.577 | 0.271 |
Chl_{t} | 346 | 28.85 | 28.80 | 15.50 | 40.65 | 4.50 | 15.59 | -0.322 | 0.013 | |
Chl_{a} | 343 | 26.65 | 26.95 | 14.85 | 31.90 | 2.33 | 8.75 | 2.600 | -0.991 | |
57 | Chl_{b} | 343 | 9.03 | 9.10 | 3.95 | 14.75 | 1.62 | 17.98 | -0.015 | -0.033 |
Chl_{t} | 343 | 35.67 | 36.05 | 20.25 | 46.65 | 3.85 | 10.78 | 0.921 | -0.507 | |
Chl_{a} | 344 | 29.91 | 30.40 | 21.05 | 47.50 | 2.62 | 8.77 | 6.333 | 0.533 | |
64 | Chl_{b} | 344 | 9.64 | 9.60 | 1.60 | 20.10 | 1.74 | 18.08 | 4.195 | 0.444 |
Chl_{t} | 344 | 39.55 | 39.93 | 26.80 | 55.20 | 3.83 | 9.69 | 1.329 | -0.014 | |
Chl_{a} | 346 | 32.16 | 32.25 | 25.35 | 37.55 | 1.93 | 5.99 | 0.405 | -0.483 | |
78 | Chl_{b} | 346 | 10.03 | 10.05 | 5.95 | 19.00 | 1.29 | 12.90 | 6.166 | 0.889 |
Chl_{t} | 346 | 42.19 | 42.60 | 32.30 | 51.30 | 2.85 | 6.76 | 0.655 | -0.500 | |
Chl_{a} | 344 | 31.23 | 31.50 | 25.20 | 34.50 | 1.54 | 4.94 | 1.369 | -0.960 | |
85 | Chl_{b} | 344 | 10.13 | 10.20 | 3.25 | 14.15 | 1.42 | 14.03 | 0.898 | -0.245 |
Chl_{t} | 344 | 41.36 | 41.43 | 32.20 | 48.55 | 2.66 | 6.42 | 0.550 | -0.550 | |
Chl_{a} | 346 | 30.12 | 30.15 | 25.65 | 34.35 | 1.56 | 5.17 | -0.049 | -0.008 | |
99 | Chl_{b} | 346 | 10.06 | 10.00 | 6.70 | 15.05 | 1.27 | 12.62 | 0.440 | 0.400 |
Chl_{t} | 346 | 40.18 | 40.05 | 32.90 | 48.70 | 2.65 | 6.60 | 0.157 | 0.110 | |
Chl_{a} | 344 | 32.62 | 32.80 | 23.15 | 38.05 | 1.87 | 5.73 | 1.813 | -0.475 | |
120 | Chl_{b} | 344 | 9.40 | 9.40 | 5.55 | 16.20 | 1.26 | 13.43 | 2.196 | 0.427 |
Chl_{t} | 344 | 42.02 | 42.30 | 29.75 | 50.50 | 2.87 | 6.84 | 0.949 | -0.394 | |
Area 2 | ||||||||||
Chl_{a} | 303 | 22.05 | 22.15 | 13.40 | 29.35 | 2.94 | 13.32 | 0.110 | -0.197 | |
41 | Chl_{b} | 303 | 6.44 | 6.50 | 2.95 | 11.15 | 1.33 | 20.70 | 0.443 | 0.234 |
Chl_{t} | 303 | 28.49 | 28.60 | 16.75 | 39.85 | 4.21 | 14.76 | 0.049 | -0.087 | |
Chl_{a} | 303 | 28.34 | 28.60 | 20.40 | 32.95 | 2.24 | 7.90 | 0.316 | -0.597 | |
57 | Chl_{b} | 303 | 8.67 | 8.55 | 5.20 | 12.50 | 1.34 | 15.43 | -0.112 | 0.128 |
Chl_{t} | 303 | 37.00 | 37.35 | 26.35 | 45.45 | 3.18 | 8.60 | 0.671 | -0.515 | |
Chl_{a} | 303 | 30.98 | 31.30 | 16.20 | 38.65 | 2.31 | 7.45 | 6.149 | -1.334 | |
64 | Chl_{b} | 303 | 9.03 | 9.05 | 4.40 | 14.00 | 1.36 | 15.07 | 1.906 | -0.008 |
Chl_{t} | 303 | 40.01 | 40.55 | 21.80 | 50.70 | 3.44 | 8.59 | 3.472 | -0.932 | |
Chl_{a} | 301 | 32.26 | 32.25 | 26.80 | 37.75 | 1.51 | 4.67 | 1.431 | 0.045 | |
78 | Chl_{b} | 301 | 10.14 | 10.05 | 5.45 | 15.80 | 1.27 | 12.49 | 1.844 | 0.463 |
Chl_{t} | 301 | 42.40 | 42.30 | 33.85 | 53.35 | 2.51 | 5.91 | 2.198 | 0.478 | |
Chl_{a} | 299 | 31.12 | 31.15 | 26.10 | 37.35 | 1.34 | 4.31 | 1.902 | 0.004 | |
85 | Chl_{b} | 299 | 9.88 | 9.85 | 7.10 | 13.45 | 1.11 | 11.19 | 0.058 | 0.189 |
Chl_{t} | 299 | 41.00 | 41.00 | 33.20 | 49.10 | 2.20 | 5.38 | 0.959 | -0.060 | |
Chl_{a} | 302 | 32.25 | 32.20 | 23.95 | 38.15 | 1.73 | 5.36 | 1.785 | -0.214 | |
99 | Chl_{b} | 302 | 11.07 | 10.95 | 6.00 | 16.95 | 1.26 | 11.39 | 2.219 | 0.482 |
Chl_{t} | 302 | 43.33 | 43.45 | 29.95 | 52.80 | 2.78 | 6.41 | 1.950 | -0.132 | |
Chl_{a} | 302 | 34.56 | 34.65 | 28.80 | 38.95 | 1.62 | 4.68 | 0.315 | -0.273 | |
120 | Chl_{b} | 302 | 9.94 | 9.80 | 5.70 | 16.50 | 1.16 | 11.67 | 3.788 | 0.807 |
Chl_{t} | 302 | 44.49 | 44.50 | 35.50 | 51.50 | 2.54 | 5.71 | 0.496 | -0.199 |
n: number of sampled plants; Med: median; Min: minimum value; Max: maximum value; s: standard deviation; CV: coefficient of variation; Ck: coefficient of kurtosis; Cs: coefficient of skewness.
The relative variability of the Chl_{a}, Chl_{b}, and Chl_{t} indices was higher in Area 1 than in Area 2 on all assessment dates. Therefore, even adjacent areas in a vineyard, assessed in the same period of the cycle, cultivated with a single grapevine cultivar, and subjected to the same agricultural practices exhibit different levels of variability regarding the relative content of photosynthetic pigments throughout the period between flowering and fruit maturation.
Moreover, in Areas 1 and 2 the variability of the Chl_{a}, Chl_{b}, and Chl_{t} indices followed a similar pattern according to the progression of the grapevine cycle. This was evidenced by the high CV values obtained in the first evaluations, soon after the onset of flowering. These values subsequently decreased and present smaller differences between each other starting from 78 DAP, before the onset of maturation. These results indicate the existence of a lower variability of photosynthetic pigment content in grapevine leaves during periods close to and after the onset of fruit maturation, a phenological stage called veraison, which was identified at 92 DAP in the assessed cycle. According to ^{Keller (2015)}, this stage is marked by the decrease in berry firmness and the increase in fruit sugar content and is followed by a quick change in fruit color, i.e., from green to red in red grape cultivars and from green to yellow in white grape cultivars, which is the case for the Chardonnay cultivar.
It was inferred that the variation in the mean values of the chlorophyll indices assessed throughout the growth cycle of Chardonnay grapevine, from the phase that follows flowering onset until maturation (Figure 2), exhibited a decreasing quadratic trend. Thus, given the distribution of the mean values of the chlorophyll indices during the assessment period, non-linear models, more specifically quadratic functions, were fitted to the temporal series because they provided a satisfactory description of the trend of the means of the three indices throughout that interval of the grapevine cycle. In addition, this type of function was selected because it was the simplest form of representing the relationship between the considered variables.
According to the analysis of variance for the regression models (Table 2), the F test confirmed the existence of a significant non-linear relationship between the chlorophyll indices measured in both areas and the phenological stages of the grapevine crop, represented by days after pruning, assuming a significance level of 5% (α = 0.05). However, after the polynomial regression equations for Chl_{a} index variation were calculated, the quadratic effect was deemed significant only for Chl_{b} and Chl_{t} indices, assuming α = 0.05. High (close to 1) coefficients of determination (adjusted r^{2}) fitted to the variation of Chl_{b} and Chl_{t} indices confirmed the models' quality of fit. Thus, most of the variation of the Chl_{b} and Chl_{t} indices in Chardonnay grapevine leaves occurs as a response to crop cycle progression and this relationship can be described with a 2^{nd} order polynomial function. Because there was no evidence that the fitted polynomial regression models were able to describe the variation trend of the Chl_{a} index with grapevine cycle development, it was assumed that chlorophyll content tended to increase in Chardonnay grapevine leaves until the considered assessment date (120 DAP), i.e. even after the onset of fruit maturation.
Index | b_{0} | b_{1} | b_{2} | Pr (>F) | r^{2} | Adjusted r^{2} |
---|---|---|---|---|---|---|
Area 1 | ||||||
Chl_{a} | 7.6383 | 0.4749 | -0.0023 | 0.0172 | 0.8690 | 0.8035 |
Pr (> t) | 0.2597 | 0.0357 | 0.0709 | |||
Chl_{b} | -2.9834 | 0.2936 | -0.0016 | 0.0037 | 0.9393 | 0.9090 |
Pr (> t) | 0.1509 | 0.0026 | 0.0040 | |||
Chl_{t} | 4.6549 | 0.7685 | -0.0039 | 0.0080 | 0.9106 | 0.8659 |
Pr (> t) | 0.5242 | 0.0118 | 0.0223 | |||
Area 2 | ||||||
Chl_{a} | 6.4299 | 0.5029 | -0.0023 | 0.0108 | 0.8963 | 0.8444 |
Pr (> t) | 0.3488 | 0.0340 | 0.0774 | |||
Chl_{b} | -1.4495 | 0.2443 | -0.0012 | 0.0017 | 0.9594 | 0.9390 |
Pr (> t) | 0.3555 | 0.0026 | 0.0052 | |||
Chl_{t} | 4.9805 | 0.7472 | -0.0035 | 0.0037 | 0.9390 | 0.9085 |
Pr (> t) | 0.4695 | 0.0102 | 0.0242 |
b_{0}, b_{1}, and b_{2}: estimated parameters of the quadratic regression model
According to the simulated values obtained using the regression equations that represent the trend lines of leaf pigment content variation over time, the highest means expected for the Chl_{b} and Chl_{t} indices in Area 1 would be obtained approximately at 92 and 99 DAP, respectively. In Area 2, the Chl_{b} and Chl_{t} indices would reach the highest simulated values at 99 and 106 DAP, respectively. In both cases, maximum values of chlorophyll b and total chlorophyll contents were estimated to occur straight after the onset of fruit maturation. In line with this finding, ^{Filimon et al. (2016)} stated that the fruit maturation phase may be considered the starting point of leaf senescence in grapevines. Other authors (^{Maia & Piedade, 2002}; ^{Zhang et al., 2008}; ^{Paula et al., 2015}) also obtained results indicating chlorophyll content variation according to the phenological stages in other plant species, in a pattern similar to that reported in that study. Specifically in wine grape cultivars, chlorophyll content variation in leaves of Pinot Noir with leaf age and position in the plant has been described by ^{Bertamini & Nedunchezhian (2002)}.
In addition to the natural pattern of variation of chlorophyll content over time, the effect of other variables on the relative chlorophyll indices was inferred from the variation of the mean values obtained in the period in question, compared with the values simulated by the trend lines fitted by the least squares method. Therefore, in addition to leaf age, plant nitrogen (N) dynamics stands out as one of the factors that may explain pigment content variation, because this element is directly related to chlorophyll content. According to ^{Arrobas et al. (2014)}, the reduction in N content throughout the grapevine cycle can also be attributed to the remobilization of this nutrient to the plant's perennial structures, as well as loss to the atmosphere. According to the previously mentioned authors, this is evidenced by the comparison between the concentration of N in chlorotic leaves and in green leaves and by the concomitant increase in N content in the woody parts of the plant.
Another possible explanation for the temporal variation in photosynthetic pigment content is the pattern of N demand by the grapevine throughout its development cycle. This demand tends to be greater between the phases of budding and flowering, with a gradual increase in the supply of N from the soil during flowering, fruiting, and the first stage of berry growth; this supply may be even greater after the onset of maturation (^{Keller, 2015}). The effect of nitrogen fertilization on chlorophyll content and the positive correlation between the relative chlorophyll index and leaf N content in grapevine cultivars was observed by ^{Tecchio et al. (2011)} and ^{Taskos et al. (2015)}. In addition, effects on the relative index of leaf chlorophyll in table grape cultivars in response to the variation in fertilizer application rate, including sources of N, P_{2}O_{5}, and K_{2}O, have been shown by ^{Andrade et al. (2017)}.
Considering only the Chl_{t} results, because this index is the sum the remaining indices, there was a modest temporal stability of total chlorophyll content in the assessed period in both areas. This was inferred from the low values of r_{s} (close to zero) obtained in the comparison between all pairs of assessment dates (Table 3). Despite the values of r_{s}, the most significant correlation occurred before the onset of fruit maturation, between 64 and 78 DAP in Area 1 and between 78 and 85 DAP in a period nearer to the veraison stage in Area 2.
Area 1 | |||||||
---|---|---|---|---|---|---|---|
41 DAP | 57 DAP | 64 DAP | 78 DAP | 85 DAP | 99 DAP | 120 DAP | |
41 DAP | 1 | ||||||
57 DAP | 0.179 ** | 1 | |||||
64 DAP | 0.137 * | 0.326 ** | 1 | ||||
78 DAP | 0.134 * | 0.160 ** | 0.335 ** | 1 | |||
85 DAP | 0.208 ** | 0.201 ** | 0.258 ** | 0.255 ** | 1 | ||
99 DAP | 0.214 ** | 0.095 ^{ns} | 0.084 ^{ns} | 0.193 ** | 0.288 ** | 1 | |
120 DAP | 0.176 ** | 0.154 ** | 0.208 ** | 0.242 ** | 0.245 ** | 0.302 ** | 1 |
Area 2 | |||||||
41 DAP | 57 DAP | 64 DAP | 78 DAP | 85 DAP | 99 DAP | 120 DAP | |
41 DAP | 1 | ||||||
57 DAP | 0.048 ^{ns} | 1 | |||||
64 DAP | 0.098 ^{ns} | 0.177 ** | 1 | ||||
78 DAP | 0.172 ** | 0.096 ^{ns} | 0.111 ^{ns} | 1 | |||
85 DAP | 0.097 ^{ns} | 0.150 ** | 0.226 ** | 0.381 ** | 1 | ||
99 DAP | 0.122 * | 0.171 ** | 0.229 ** | 0.209 ** | 0.255 ** | 1 | |
120 DAP | -0.034 ^{ns} | 0.084 ^{ns} | -0.003 ^{ns} | 0.068 ^{ns} | 0.047 ^{ns} | 0.060 ^{ns} | 1 |
Results of the statistical significance test:
^{*}significant at 5% (p < 0.05);
^{**}significant at 1% (p < 0.01); ns: non-significant.
The analysis of the spatial variation of the Chl_{t} index showed different patterns of distribution of total chlorophyll content according to DAP. As is shown in Table 4, the model that best fitted the majority of experimental semivariograms was the exponential model, whereas the spherical model was only used for Chl_{t} data of the measurements made at 41 DAP (Area 1) and at 78 and 85 DAP (Area 2). The geostatistical analysis of the Chl_{t} data showed the presence of geometric anisotropy (^{Oliver & Webster, 2014}) only in Area 1 at 64, 78, and 85 DAP, with longer range of spatial dependence (greater continuity) in the 75°, 90°, and 90° directions, respectively.
DAP | Model | Nugget effect | Level | Range(m) | SDI (%) | RMSE |
---|---|---|---|---|---|---|
Area 1 | ||||||
41 | Spherical | 18.09 | 20.66 | 55.10 | 3.61^{W} | 4.39 |
57 | Exponential | 9.71 | 14.89 | 5.20 | 0.80^{W} | 3.93 |
64 | Exponential | 7.74 | 16.25 | 20.00 | 4.66^{W} | 3.77 |
78 | Exponential | 4.57 | 9.36 | 45.78 | 10.42^{M} | 2.64 |
85 | Exponential | 4.32 | 7.68 | 24.30 | 4.73^{W} | 2.55 |
99 | Exponential | 3.52 | 7.23 | 21.90 | 4.99^{W} | 2.45 |
120 | Exponential | 4.23 | 8.27 | 16.59 | 3.61^{W} | 2.82 |
Area 2 | ||||||
41 | Exponential | 9.21 | 18.15 | 6.56 | 1.98^{W} | 4.28 |
57 | Exponential | 7.34 | 10.26 | 17.46 | 3.05^{W} | 3.13 |
64 | Exponential | 6.74 | 9.83 | 12.02 | 2.32^{W} | 3.12 |
78 | Spherical | 0.16 | 6.01 | 5.20 | 3.68^{W} | 2.64 |
85 | Spherical | 1.96 | 4.65 | 5.62 | 2.36^{W} | 2.29 |
99 | Exponential | 2.64 | 7.49 | 5.83 | 2.32^{W} | 2.76 |
120 | Exponential | 3.07 | 6.32 | 6.11 | 1.93^{W} | 2.56 |
SDI: spatial dependence index (^{Seidel & Oliveira, 2014}); W, M, and S: weak, moderate and strong spatial dependence, respectively (^{Seidel & Oliveira, 2016}). RMSE: root mean squared error.
In Area 1, the variation in the range of spatial dependence of the data was greater when the two first assessment dates were compared, with variogram amplitude results of 55.10 and 5.20 m, respectively. In the remaining assessment days in the same area, the range of spatial dependence was between 16.59 and 45.78 m. This indicated that spatial dependence of the Chl_{t} data in Area 1 was more stable starting from 64 DAP, i.e. there was a consolidation of the spatial continuity of chlorophyll content represented by the Chl_{t}, a result that is in line with the coefficient r_{s} (Table 3).
In Area 2, the values of maximum range of spatial dependence on all assessment dates were lower than those of the semivariograms fitted to the sets of data for Area 1, varying between 5.20 and 17.46 m at 78 and 57 DAP, respectively. Given the magnitude of the values of this parameter, it was inferred that using measurements made in a higher sampling density would improve spatial inference of the Chl_{t} index in Area 2, because the range of spatial dependence represented the distance at which maximum variability was identified (^{Grego & Oliveira, 2015}).
In the semivariograms for Area 1, spatial dependence of the Chl_{t} index was classified as weak on all assessment dates, with the exception of 78 DAP when it was considered moderate. Spatial dependence associated with the semivariograms of the Chl_{t} index measured in Area 2 was classified as weak on all assessment dates. Because the results of SDI were low throughout the evaluated period, it was not possible to identify the phase of the cycle in which the results of kriging, and therefore of Chl_{t} index mapping, had the lowest uncertainty. Thus, this result showed the heterogeneity of the spatial distribution of total chlorophyll content in Chardonnay grapevines within the same vineyard. Despite the results obtained for the degree of spatial dependence of the Chl_{t} index, the validation of the semivariogram fitting models was confirmed by the value of the root-mean-squared error (RMSE) in all assessment dates, with the lowest values obtained at 99 DAP (2.55 in Area 1) and at 85 DAP (2.29 in Area 2).
Figures 3 and 4 show the thematic maps of spatial distribution of the Chl_{t} index in Areas 1 and 2, respectively, based on the classification of the interpolated values of Chl_{t} into three categories. In the maps, regions with dark colors show higher estimated values of Chl_{t} (Class 1), whereas lighter colors indicate sites with lower values of Chl_{t} (Class 3), according to the values obtained on each assessment date.
The comparison of the categorical maps of the Chl_{t} index obtained on subsequent assessment dates (Table 5) showed that values of V_{1} close to zero obtained in both areas and intervals indicated a weak spatial association between the regionalization of total chlorophyll content. Low values of GOF also indicated a modest degree of spatial association between the pairs of categorical maps of the Chl_{t} index estimated throughout the assessment period. However, considering the two mentioned criteria, the strongest associations were observed between the thematic maps for the intervals between 64 and 78 DAP (Area 1) and between 78 and 85 DAP (Area 2), which were shown to be more similar, a finding that is in line with the higher temporal stability estimated using Spearman's rank correlation analysis (Table 3).
Area 1 | ||||||
---|---|---|---|---|---|---|
Parameter | 41-57 DAP | 57-64 DAP | 64-78 DAP | 78-85 DAP | 85-99 DAP | 99-120 DAP |
V_{1} | 0.01 | 0.01 | 0.17 | 0.09 | 0.13 | 0.04 |
GOF | 0.35 | 0.35 | 0.45 | 0.41 | 0.44 | 0.36 |
Area 2 | ||||||
Parameter | 41-57 DAP | 57-64 DAP | 64-78 DAP | 78-85 DAP | 85-99 DAP | 99-120 DAP |
V_{1} | 0.01 | 0.02 | <0.01 | 0.04 | 0.02 | 0.01 |
GOF | 0.34 | 0.35 | 0.33 | 0.35 | 0.34 | 0.35 |
V_{1}: V-measure; GOF: goodness of fit.
In Area 1, the similarity between the Chl_{t} index maps was lower in the comparison of the three first pairs of assessments than in the comparison of the following assessments, and it decreased again at the end of the period in question. From 64 DAP until 85 DAP, there was a concentration of high and intermediate values (Class 1 and Class 2) of Chl_{t} index in the central region and a concentration of low values (Class 3) in the upper and lower limits, which resulted in similar patterns of Chl_{t} index spatial distribution in this period. Thus, in Area 1, the spatial distribution of the classes of this attribute varied less during these intervals than in the remaining intervals.
According to the V_{1} and GOF values obtained for Area 2, the similarity between the maps was low and stable throughout the assessed period, given the pattern of spatial distribution of the regionalization of the Chl_{t} index, characterized by the instability of the layout of the classes of total chlorophyll content in the area and of the variation of the patterns obtained on each occasion. Moreover, although it was possible to identify classes of Chl_{t} index in this vineyard, they exhibited excessive dispersion and fragmentation resulting in smaller adjacent areas, which may preclude the use of this spatial analysis for the specific management of the area.
Considering only Area 1, the categorical map of Chl_{t} index obtained between 64 and 78 DAP and between 85 and 99 DAP, especially, may serve to guide the use of the portable meter for monitoring the nutritional status of the plant regarding N content in regions of interest within the vineyard. This is possible because the measurement given by this instrument correlates with leaf N content. These results were also related to the determination of the optimal phase for leaf sampling for the inference of the nutritional status of the grapevine, because the sampling timing affects the diagnosis of the concentration of mineral nutrients in the plant (^{Arrobas et al., 2014}). In grapevine crops, it is more appropriate to perform nutritional assessment in samples of leaves collected during full flowering and at the onset of fruit maturation. Thus, assuming that the onset of maturation in the evaluated cycle occurred at 92 DAP, spatial analysis of the Chl_{t} index performed in the interval between 64 and 99 DAP could have been used to delimit more regular sampling regions regarding total chlorophyll content in Chardonnay grapevine leaves. The determination of leaf N content performed through the measurement of Chl_{t} index (according to its spatial variability and during the adequate period) together with soil fertility analysis may be useful to define areas for variable rate fertility management, thereby increasing the efficiency of fertilizer recommendations in the vineyard.
CONCLUSIONS
The mean variation of the Chl_{b} and Chl_{t} indices in Chardonnay grapevine leaves throughout its cycle may be represented by the quadratic trend used to estimate the reduction in the content of chlorophyll b and total chlorophyll after the onset of fruit maturation. In contrast, high mean values of Chl_{a} index are still detected after the onset of this stage.
The Chl_{t} index of Chardonnay grapevine exhibited a low temporal and spatial stability in the period between flowering and full fruit maturation in geographically close vineyards under similar cultivation and management conditions. However, the period that preceded the onset of fruit maturation was associated with a higher temporal stability of indirect measures of total chlorophyll content, with a higher similarity between consecutive categorical mappings of this variable.