Improving bitter pit prediction by the use of X-ray fluorescence (XRF): A new approach by multivariate classification

Bitter pit (BP) is one of the most relevant post-harvest disorders for apple industry worldwide, which is often related to calcium (Ca) deficiency at the calyx end of the fruit. Its occurrence takes place along with an imbalance with other minerals, such as potassium (K). Although the K/Ca ratio is considered a valuable indicator of BP, a high variability in the levels of these elements occurs within the fruit, between fruits of the same plant, and between plants and orchards. Prediction systems based on the content of elements in fruit have a high variability because they are determined in samples composed of various fruits. With X-ray fluorescence (XRF) spectrometry, it is possible to characterize non-destructively the signal intensity for several mineral elements at a given position in individual fruit and thus, the complete signal of the mineral composition can be used to perform a predictive model to determine the incidence of bitter pit. Therefore, it was hypothesized that using a multivariate modeling approach, other elements beyond the K and Ca could be found that could improve the current clutter prediction capability. Two studies were carried out: on the first one an experiment was conducted to determine the K/Ca and the whole spectrum using XRF of a balanced sample of affected and non-affected ‘Granny Smith’ apples. On the second study apples of three cultivars (‘Granny Smith’, ‘Brookfield’ and ‘Fuji’), were harvested from two commercial orchards to evaluate the use of XRF to predict BP. With data from the first study a multivariate classification system was trained (balanced database of healthy and BP fruit, consisting in 176 from each group) and then the model was applied on the second study to fruit from two orchards with a history of BP. Results show that when dimensionality reduction was performed on the XRF spectra (1.5 - 8 KeV) of ‘Granny Smith’ apples, comparing fruit with and without BP, along with K and Ca, four other elements (i.e., Cl, Si, P, and S) were found to be deterministic. However, the PCA revealed that the classification between samples (BP vs. non-BP fruit) was not possible by univariate analysis (individual elements or the K/Ca ratio).Therefore, a multivariate classification approach was applied, and the classification measures (sensitivity, specificity, and balanced precision) of the PLS-DA models for all cultivars evaluated (‘Granny Smith’, ‘Fuji’ and ‘Brookfield’) on the full training samples and with both validation procedures (Venetian and Monte Carlo), ranged from 0.76 to 0.92. The results of this work indicate that using this technology at the individual fruit level is essential to understand the factors that determine this disorder and can improve BP prediction of intact fruit.

1 Introduction

Bitter pit (BP) is considered one of the most relevant postharvest disorders in the apple industry worldwide (Al Shoffe et al., 2019). It occurs on a wide geographic range and a significant number of cultivars (Bergmann, 1992; Volz et al., 2006), causing major economic damage to growers and exporters, as it develops progressively during storage (Lötze and Theron, 2006). BP appears mainly in the subepidermal tissue at the calyx end of the fruit, as internal corky lesions (Jarolmasjed et al., 2016; Jemrić et al., 2016). It was first described 150 years ago (Oberdieck and Lucas, 1869), and remains as one of the most studied disorder of apples, except for scald. In general terms, BP is associated with a localized calcium (Ca) deficiency (Perring and Pearson, 1986; Ferguson and Watkins, 1992; Saure, 2005; Fallahi et al., 2006; Fallahi et al., 2010; Sharma et al., 2012; Espinosa-Zúñiga et al., 2017; Torres et al., 2017) in association with an imbalance of other minerals, most frequently nitrogen (N), potassium (K) and magnesium (Mg) (Smock and Van Doren, 1937; Perring and Pearson, 1986; Fallahi et al., 1997; Fallahi et al., 2006; Fallahi et al., 2010; Jarolmasjed et al., 2016; Jemrić et al., 2016; Kalcsits, 2016; Al Shoffe et al., 2019; Kalcsits et al., 2019; Torres et al., 2021). Phosphorus (P), like Ca and Mg, have been found to accumulate in pitted tissue (Chamel and Bossy, 1981).

Ca reaches the roots by mass flow (Jungk and Claassen, 1997) and is translocated to the end of the vascular tissues (shoots, leaves and fruits) by transpiratory flow throughout the xylem (Medrano et al., 2007; López López et al., 2009). Exchangeable soil Ca is rarely responsible for the occurrence of BP, at least not in temperate soils, where it yields more than 50% of exchangeable soil cations (Ca, Mg, K and Na) (Blume et al., 2016). Thus, Ca uptake and subsequent Ca deficiency in fruit is not necessarily a consequence of soil Ca restriction. Consequently, soil applications of Ca have not been as effective in reducing BP in apples as foliar applications (Torres et al., 2017), although in apple orchards planted in acid soils treated with lime or gypsum, a decrease in BP has been observed (Wilms and Basso, 1987). As the fruit develops, the contribution of the xylem to the total inflow to the fruit decreases and the phloem flow increases, implying less Ca delivery to the fruit (Lang, 1990). This loss of xylem functionality, due to the fruit growth, disrupts water mass flow, limiting the transport of Ca to the fruit, which has been suggested to favor development of BP (Alarcón et al., 1998; Miqueloto et al., 2014). Due to this, prediction of BP occurrence by mineralogical analysis in fruit apples is carried out between 40 and 60 days after full bloom, before Ca is diluted up to 50% by fruit growth (Peryea et al., 2007; Miqueloto et al., 2014).

Several factors predisposing and intensifying nutritional imbalances at the limb/fruit level have been associated with BP severity. For example, excessive tree vigor (Terblanche et al., 1979; Baugher et al., 2017) due to an inappropriate rootstock/cultivar combination in response to both soil (Sió et al., 1999; Weibel et al., 2000; Biskup et al., 2003; Fallahi, 2012; Fazio et al., 2013; Sió et al., 2018) and the environmental characteristics (Goode and Ingram, 1971; Bergmann, 1992; Jemrić et al., 2016), as well as the agronomic management that impacts on plant light interception (Van Der Boon, 1980), nutrition (Fallahi et al., 1997; Kim and Ko, 2004), crop load (Ferguson and Watkins, 1992; Volz et al., 1993; Tough et al., 1998; Volz and Ferguson, 1999; Telias et al., 2006; Seo et al., 2007), and fruit ripeness at harvest (Al Shoffe et al., 2020).

Since BP damage often becomes evident after several months of refrigerated storage (Conway et al., 2002; Jemrić et al., 2016) and control methods are not always effective (Fernández et al., 2009), significant efforts have been committed to developing methodologies for early prediction of the disorder. Fruit mineral analysis and infiltration by Mg salts remain the most used predictive tools, but with different levels of effectiveness (Baugher et al., 2017). The former, is used to estimate Ca content as well as its relationship with other nutrients. In general this method has a low predictive capacity (10 – 40%) given the need to compose the sample with more than one fruit (Al Shoffe et al., 2019). Nevertheless, in a study where the mineral concentration of the fruit was determined three weeks before commercial harvest, it was concluded that (K+Mg)/Ca, (N+K+Mg)/Ca or N/Ca ratios were highly correlated with the occurrence of BP in ‘Honey Crisp` apples, which is a very susceptible cultivar (Marini et al., 2020). The main disadvantage is the analysis based on a group of fruits, that increases variability on the results. On the other hand, infiltration by Mg salts has a more significant association with BP (<70%), but is time consuming, destructive and difficult to implement massively (Bangerth, 1974; Hopfinger and Poovaiah, 1979; Burmeister and Dilley, 1991; Retamales et al., 2000; Retamales et al., 2001; Amarante et al., 2010; Torres et al., 2015).

The relevance of the relationship between K and Ca is well known, and is of particular interest for the present study. For example, Val et al. (1999) reported that the risk of BP increased considerably when, between 80 – 100 d after full bloom, the ratio in fruit increased above 25 and in leaf fell below two. Today, K/Ca remains as a reference for any cultivar anywhere in the world (Lötze et al., 2008; von Bennewitz et al., 2015; Espinosa-Zúñiga et al., 2017). Recently, Prengaman (2021) proposed, as indicator of BP occurrence, the nutrient content of apple juice, indicating that K/Ca ratio explained about 70% of the variation in BP of ‘Honey Crisp’ samples. Although the K/Ca ratio could provide valuable information for the industry, a high variability of Ca and K content has been reported within the fruit as well as between fruits on a plant (Wills et al., 1976; Ferguson and Triggs, 1990; Le Grange et al., 1998; Lötze and Theron, 2006).

Therefore, it is pertinent to study the problem at the individual sample level and, ideally, using non-destructive tools. Although some non-destructive approaches are not able to determine the nutrient content (i.e. non-mineral approach), they can estimate it. In this sense, proximal and non-proximal remote sensing equipment, such as VIS/NIR/SWIR spectroscopy (i.e., ~350 – 2,500 nm) are widely used to estimate the nutritional status of plant tissues (Nicolaï et al., 2006; Kafle et al., 2016; Jarolmasjed et al., 2017; Jarolmasjed et al., 2018; Mogollón et al., 2021), mainly in leaves but also in other organs (García-Sánchez et al., 2017). For example, spectral reflectance at fruit level was useful in identifying BP-fruit, however, other lesions were misclassified as BP (Nicolaï et al., 2006; Jarolmasjed et al., 2017).

Less developed in fruit research, X-ray fluorescence (XRF) spectroscopy was designed for the semi-quantitative measurement of mineral levels in different materials, thus with potential to determine mineral elements and their ratios (e.g., K/Ca) (Espinosa-Zúñiga et al., 2017). Handheld XRF devices accurately quantify the photon emission of elements with an atomic weight greater than 28.1 (i.e., silicon, Si), which includes Ca and K. Kalcsits (2016) demonstrated that XRF could be used for non-destructive semi-quantitative determination of Ca and K in apples and pears, and there was a significant high correlations with mineralogical content coming from traditional laboratory analysis; however, like others (Mohr and Jamieson, 1984; Baugher et al., 2017), this work highlights that beyond K/Ca, other elements are implicated in the occurrence of the disorder. Therefore, it is hypothesized that using a multivariate modeling approach, other elements in addition to K and Ca could be found to improve the current predictive capability of BP.

2 Material and methods

2.1 Apple fruit used for bitter pit prediction

Two studies were carried out as follows:

2.1.1 Study 1

A balanced database was generated with apples with and without BP symptoms. For this purpose, during the 2018/19 season, Dole Chile Co. provided two groups of ‘Granny Smith’ apples from a commercial lot of refrigerated storage (5 months at 0°C and 90% RH): i) lot 1 with BP: 176 apples with medium and severe epidermal incidence of BP (i.e., 3 – 5, and >5 pits, respectively); and ii) lot 2 without BP: 176 apples with no evidence of BP injury (i.e., no epidermal indication of BP or other disorder). Since most of the BP damage is concentrated towards the distal region of the fruit, the evaluations were concentrated in the calyx area (Figure 1A-1). To determine the variability associated with XRF measurements at fruit level, a reproducibility study was conducted, in which six equidistant points were measured along the calyx end of each pitted and non-pitted fruit (Figure 1A-2).

FIGURE 1

Figure 1 Section of the fruit calyx (dashed red line) where the bitter pit is usually located (A-1) and where the reproducibility XRF study was performed at six equidistant points (A-2); details of XRF measurement (B); bitter pit in fruit with no external symptoms, before (C-1) and after fruit peeling (C-2).

2.1.2 Study 2

To predict the occurrence of BP, apples harvested from a commercial orchard were used. The fruit were collected during the 2019/20 season from two orchards with previous records of medium to high BP incidence. The orchards are located in the Maule Region – Chile: i) San Clemente (35°30’52” S; 71°26’42.8” W): ‘Granny Smith’/MM9 and ‘Brookfield’/MM9, planted in 2011 (3.5 x 1.2 m; NW/SE orientation); and ii) Lontue (35°15’39” S; 71°14’32.0” W): ‘Fuji’/MM9, planted in 2007 (4.5 x 1.5 m; NW/SE orientation). Trees were trained under Solaxe system, with three distinguishable productive floors, having the upper third a lower amount of fruit and leaves compared to the inferior floors (i.e., lower and middle); the insertion of each branch into the trunk determined the floor to which fruit corresponded. Given that, in previous seasons, more than 70% of the BP (developed after storage) was found in the upper floor of the trees in these orchards (data not shown), during the 2019/20 season, all fruit from the upper floor of ten homogeneous trees of each orchard and cultivar, were collected at harvest. The fruit were transferred to the Postharvest Laboratory of the Universidad of Talca where elemental composition was determined with a handheld XRF device by measuring the calyx area of each fruit (Figure 1). After this, apples were stored for 5 months at 0°C and 90% RH to induce BP appearance. Upon storage removal, fruit were placed at room temperature at 20°C for 10 d to accelerate the expression of the disorder. The incidence and severity of BP were visually scored and classified (Figure 1B) as undamaged (i.e., without evident damage inside and outside the fruit) and damaged fruit (i.e., mild, medium, and severe BP, consisting of<2; 3 – 5, and >5 pits, respectively), by peeling all fruit at the calyx end.

The operation condition of the XRF instrument (Titan S1, Bruker Elemental portable handheld X-ray fluorometer, Kennewick, WA, USA) was a voltage of 21 kV, amperage of 40 µA, and 15 s of exposure. As proposed by Kalcsits (2016) to ensure that as many X-rays bombard the sample, each apple was placed with as much contact as possible with the instrument’s surface (Figure 1C). Since the device is sensitive to X-rays with photon energies above 1.5 keV, it is capable of simultaneously determining chemical elements heavier than silicon (Si). Consequently, elements other than K and Ca were studied at the same time; between 1.5 and 8 keV. The estimated penetration depth of X-rays is approximately 1 mm, and therefore, the elemental composition of skin and flesh was determined (Kalcsits, 2016).

2.2 Statistical data analysis and modelling of the XRF signature

2.2.1 Ionomics characterization of ‘Granny Smith’ fruit with BP (Study 1)

For the exploratory analysis of the complete XRF spectra, a Principal Component Analysis (PCA) was carried out. The main objective of PCA is to reduce many variables to a smaller set of factors, named principal components (PCs) while retaining most of the information from the original dataset (Wold et al., 1987; Bro and Smilde, 2014). Briefly, the data matrix X (with n samples and m variables) was decomposed according to equation (1), where T_(n,A) is the score factor matrix and P_(m,A) is the loading factor matrix for A components.

$\begin{array}{ll}{X}_{\left(n,m\right)}=T_{\left(n,A\right)}·P_{\left(A,m\right)}^T& \text{Equation (1)}\end{array}$

In this model, the scores represent the projection of n samples on the reduced subspace of A dimensions, where A< m. The critical parameters to be adjusted are the number of principal components (PCs) and the preprocessing data analysis. In this study, the number of PCs were determinate by full-cross validation strategy and the preprocessing procedures considered were mean-centering and autoscaling of spectral data.

For a quantitative understanding of such differences between both BP groups the signal intensities of each element were compared by ANOVA, cumulative frequency distribution and box-and-whisker plots. For the particular case of K/Ca, two approaches were developed: i) deconvolution of the K and Ca peaks and subsequent calculation of the ratio (Kalcsits, 2016); and ii) simple ratio, where the values measured by the equipment were used directly (i.e., K and Ca without deconvolution).

2.2.2 Prediction of BP by multivariate XRF signature modeling (Study 2)

For classification purposes of the fruit coming from orchards with previous records of medium to high BP incidence, Partial Last Squares Discriminant Analysis (PLS-DA) was considered. The principal advantage of this method is to combine the dimensionality reduction and discriminant analysis into one algorithm, being especially useful to analyze high dimensional data (Lee et al., 2018). Basically, a PLS-calibration model is built between one X_(n,m) data matrix and the Y_(n,1) vector containing the class labels of samples. The X_(m,n) data matrix is decomposed through the PLS algorithm that searches for latent variables (LVs) with a maximum covariance with the Y-variable. Since predicted values are quantitative, then samples can be classified according to a rule, such as the labelling to the class corresponding to the highest prediction. More detailed and formal description of the algorithm is presented in a previous publication (Barker and Rayens, 2003).

All data were analyzed by using commercial software (Matlab, Mathworks Inc., MA, USA). For PCA and PLS-DA, procedures were calculated by means of Guide to User Interface (GUI) available in internet (https://michem.unimib.it/, last revision: 30 March 2022) and written in MATLAB language (Ballabio and Consonni, 2013; Ballabio, 2015).

Prior to supervised classification, data were explored by means of PCA to remove anomalous samples, which were identified based on extreme values of Q residuals and Hoteling T² (Bro and Smilde, 2014). The retained spectra were subjected to PLS-DA models on autoscaling data and applied to classification of samples classified in two groups: affected and non-affected by BP. The best classification models were obtained when fruit on the affected group excluded the mild category (i.e., only apples classified as moderate and severe incidence were considered).

The performance to distinguish BP and non-BP fruits was evaluated using the following classification measures (Ballabio et al., 2018): sensitivity (Sn, ratio of correctly classified non-BP samples), specificity (Sp, ratio of correctly classified BP samples), and balanced accuracy (BA), which is the average of sensitivity and specificity. Classification models were validated through two different approaches: i) internal cross-validation (5 groups split with the venetian blinds procedure); and ii) Montecarlo validation based on 100 iterations (in each iteration 20% of samples were randomly selected and used to test the classification models).

3 Results

3.1 Study 1: Determination of the K/Ca ratio by XRF on fruit with and without bitter pit

When the balanced data base of ‘Granny Smith’ apples with and without BP was studied, the X-ray spectra (Figure 2) showed that the most significant peaks corresponded to potassium (K: 3.31 and 3.59 keV), calcium (Ca: 3.69 and 4.01 keV), and chloride (Cl: 2.68 and 2.82 keV); less intense peaks were found for silicon (Si: 2.02 and 2.46 keV), phosphorus (P: 2.02 and 2.14 keV), and sulfur (S: 2.31 and 2.46 keV). For these spectra, the most affected samples exhibited higher K intensity, while non-affected apples showed higher Ca intensity (Figure 2).

FIGURE 2

Figure 2 General XRF spectra for ‘Granny Smith’ apple epidermis for fruit with (red) and without (blue) bitter pit; K and Ca peaks details in small graphic. Similar spectra were found in the other cultivars.

Effectively, the analyzed spectra by a PCA model showed that K (3.31 keV) and Ca (3.69 keV) peaks contributed with the maxima variability for each component (Figure 3A). The dispersion of these two signals, evaluated as standard deviation and coefficient of variation, are shown in Supplementary Table 1. The results evidenced that the random error for K and Ca responses varied in the range of 10 to 25% (data not shown); considering that instrument variation is lower than 10%, this dispersion could be associated to heterogeneity within the apple composition. Additionally, the score plot (Figure 3B) obtained after PCA analysis shows a severe overlapping between samples with and without BP, suggesting an important similarity between spectral signature of the fruits.

FIGURE 3

Figure 3 Loading (A) and score (B) plots obtained with XRF-spectra of ‘Granny Smith’ apples with (blue) and without (red) bitter pit.

Especially for the K/Ca ratio calculated after the element deconvolution of the data matrix of ‘Granny Smith’, the results showed a high variability in sound fruits (~5 – 23), differences that almost doubled in the case of affected ones (the ratio ranged from ~4 – 41) (Supplementary Table 1). Globally, 80% of the fruit without BP had a K/Ca ratio of less than ~7, while in those with BP, for the same ratio, the proportion was less than 24% (Figure 4A); although with different scales, similar results were found for K/Ca (Supplementary Table 1, and Figure 4B). From same data base, the analysis of variance (Supplementary Table 1) the cumulative frequency distribution and the box-and-whisker plots (Figures 4C–H) of all potential identified elements indicated that, except for Cl and partially for P (p=0.0931), the rest of the elements differed between BP groups (p≤ 0.00001). On the other hand, from a methodological point of view, no differences were found among the six equidistant points at the calyx-end for any of the elements demonstrating that measurements at the calyx is consistent. Furthermore, no significant interactions were found between the two factors, indicating that the main difference between samples was due to the BP presence/absence (Supplementary Table 1). K/Ca ratio significance was the same regardless of the way in which they were calculated (deconvolution method vs direct ratio from maximum K and Ca signals); also, a high association (R² = 0.98) was found between both (Supplementary Figure 1).

FIGURE 4

Figure 4 Comparison of the cumulative frequency distributions and box-and-whisker plots of K/Ca ratio calculated by the deconvolution method (decon) (A) and direct ratio (dr) (B) from maximum K and Ca signals, and the signal insensitivity of the mineral elements with the greatest preponderance in ‘Granny Smith’ fruit with (red) or without (green) bitter pit: Si (C), P (D), S (E), Cl (F), K (G), and Ca (H).

The best PCA models were obtained after mean-centering and using 4 or 5 components to reach explained variances higher than 90%. As observed in Figure 3, the score plot resulted in 69.8% of explained variance for PC1 and 14.1% for PC2 (Figure 3A) and ~90% when PC3 was included. The more explicative spectral region considers the characteristic region of K, Ca, and Cl (i.e., 2.6 to 3.7 keV). Nevertheless, a significant influence is observed for lower energy signals between 1.7 and 2.5 keV, typical for Si, S, and P.

3.2 Study 2 (field trials): Multivariate classification of bitter pit disorder by using XRF spectra on three apple cultivars

All fruit from the upper section of the tree was harvested and initially measured by XRF; from these samples BP developed on 18.4, 3.6, and 2.4% of ‘Granny Smith’, ‘Brookfield’ and ‘Fuji’ apples, respectively (average of ten trees), as detailed on Table 1. For the XRF analyses, the spectra of the apples harvested in the field (‘Brookfield’ and ‘Fuji’) (Supplementary Figure 2) were similar to those of the records generated by the ‘Granny Smith’ balanced database (section 3.1), so they would provide similar information and therefore be consistent with the objectives of the multivariate modeling.

TABLE 1

Table 1 Number of total fruits used for ‘Granny Smith’-GS, ‘Brookfield’-BF and ‘Fuji’-FJ apples coming from commercial orchards for BP prediction.

In the same way as described for ‘Granny Smith’ apples on section 3.1, the classification between samples (BP vs. non-BP fruit) was not possible by univariate analysis (individual elements or the K/Ca ratio). Therefore, multivariate approaches, using a selected spectra region (1.5 to 4.95 keV), were evaluated. A preliminary estimation showed that the classification of the different damage levels on apple samples was not possible by using linear (LDA, SIMCA and PLS-DA) and nonlinear (QDA and NNA) multivariate models due to severe overlapping of the different classes. For this reason, the classification between two groups: damaged- (including moderated and severe damage) vs. undamaged-fruit was considered for further analysis. This approach allowed to increase the balance of data set (samples with severe BP are less abundant) and improved the evaluated classification models. From the several multivariate tested methods, the best results were obtained with PLS-DA. For this, auto scaled preprocessing and 3 to 4 components were required to explain over 80% of the variance. The PLS-DA-scores obtained for ‘Granny Smith’ suggest a better separation between sound and affected fruit (Figure 5A). In addition, the higher PLS-DA-regression coefficients are obtained for first 300 channels, corresponding from 1.5 up to 4.95 keV (Figure 5B).

FIGURE 5

Figure 5 Score (A) and normalized intensity (B) plots for ‘Granny Smith’ apples coming from field trials, comparing non affected (blue) and affected (red) fruit.

As expected, the more explicative spectral region considers the characteristic section of K, Ca, and Cl (i.e., 3.0 to 4.0 keV). Nevertheless, a significant influence is observed for lower energy signals between 1.5 and 2.5 keV, typical for Si, S, and P. The same consideration can be extended to the other cultivars because similar results were obtained. Sensitivity, specificity, and balanced accuracy of PLS-DA models for all cultivars, evaluated on the full training samples and with both validation procedures (venetian blinds and Montecarlo), ranged from 0.76 to 0.92 (Table 2).

TABLE 2

Table 2 Classification measures (balanced accuracy, specificity, and sensitivity) obtained in fitting, cross validation (with 5 groups split in Venetian blinds and Montecarlo approach) for the three apple cultivars (‘Granny Smith’-GS, ‘Brookfield’-BF, and ‘Fuji’-FJ).

4 Discussion

Establishing a fruit orchard implies many agronomic decisions (e.g., site selection, rootstock, cultivar, planting density, row orientation, training system, irrigation system) that, along with practical managements (nutrition, crop load, pruning, pollination), will affect both, the vegetative and reproductive expression of the plant; thus, also the incidence of physiological disorders as BP. In the orchards of the present experiment, the BP was found mainly on the upper production floor, where there was a lower amount of fruit and leaves compared to the lower floors.

In older orchards, with planting frames such as the one studied (i.e., 3.5 – 4.5m x 1.2 – 1.5 m), if there is no reasonable control of vegetative expression, it is common to find an excess of leaves and an uneven distribution of the leaf/fruit ratio within the plant. Since Ca movement within the plant occurs through transpiration flow, the number and distribution of leaves within the canopy should also influence the location and accumulation of the element. Thus, because microclimatic characteristics at each plant floor or branch section vary between days (e.g., maximum water pressure deficit, cloudiness, and incident radiation) and throughout the season (e.g., translational effects of the sun), the location of the clutter is difficult to predict (Baugher et al., 2017). In the present study, the upper part of the tree would be subjected to a more prolonged daily radiative stress than in the lower floors, making Ca supply to that section of the tree even more challenging, thus predisposing its fruits to develop the disorder.

In the present study for all cultivars, the strongest signals in the XRF spectrum (Figure 2) corresponded to K and Ca. Although the K/Ca ratio is the most cited proxy related to the disorder, BP has been characterized as an alteration of a more complex ionic content of the damaged tissues, affecting other elements that are involved in essential physiological processes (Simons and Chu, 1980; Chamel and Bossy, 1981; Garrec, 1983; Baugher et al., 2017). Probably because of that, quite often, predictions based on such ratio can be erratic, as lots with relatively high Ca concentration (above 5 mg Ca 100 g^-1 FW) may end up developing BP, whereas fruit below that threshold don’t always do (Terblanche et al., 1980; Le Grange et al., 1998; Lötze and Theron, 2006). The range of values on the elements found in fruit with and without BP (both as cumulative frequency distributions and box-plots, Figure 3), also evidenced this trend, indicating that classification of the disorder is not possible using an univariate approach (Figure 3B). Nevertheless, because of all above discussed, more than the known relevance of the relationship between K and Ca (i.e., fruits with BP symptoms use to have higher K/Ca than those without BP), there is a need to study the ionomics of the disorder, but at the individual fruit level.

In this sense, along with K and Ca, other four elements proved to be involved in the occurrence of BP (i.e., Cl, Si, P, and S) (Figure 3A). As it can be seen from PCA the first component (i.e., K, Ca, and Cl peaks) indicates that all the samples with severe BP show a major content of these elements. In contrast, the second component has a negative influence of Cl and a positive influence of Ca and K. In addition, the influence of less abundant elements, such as Si (1.74 and 1.83 keV) and S (2.31 and 2.46 keV) appear significant for the second and third components, suggesting a more complex relationship between BP and fruit mineral composition. Similar, Chamel and Bossy (1981) reported a high variability of P and Cl in apples with and without BP, and a lower impact of S. Likewise, Ca, K, Mg, and P concentrations were higher in those apples that presented the disorder (Garrec, 1983; Al Shoffe et al., 2014). Baugher et al. (2017) developed multiple regression models for BP prediction in ‘Honey Crips’ apples, indicating that the best two-variable-model included N/Ca ratio and shoot length (R² = 0.68); when P, boron (B), S, Ca, and Mg/Ca were added to the model, the coefficient of determination improved (R² = 0.71). More recently, Marini et al. (2020) evaluated different models to predict BP on 17 `Honeycrisp´ apple orchards, concluding that the two-variable model including B and Mg/Ca ratio was the best (R^{2 =} 0.83), however, it seems to be not conclusive since in previous season the best model containing N/Ca ratio and the shoot length underestimate the BP incidence.

Despite the ability of the multivariate PCA model to identify Si, P, S, Cl, K and Ca as potential elements involved in the disorder, there was no clear differentiation (i.e., strong data overlapping) between BP severity samples (none, moderate and severe symptoms, Figure 3B). Nevertheless, when multivariate classification approaches (i.e., contrasting fruit with vs. without BP) were considered, discrimination between sound and BP affected fruit improved significantly; for all cultivars, the sensitivity and specificity were higher than 0.76 for the training set, demonstrating that the models allowed to correctly recognize at least 78% of the samples affected by BP and reject the class “damaged” for at least the 76% of non-affected samples. In addition, the balanced accuracy showing a similar result, evidenced that the calibration model classifies correctly over 77% of analyzed samples.

Finally, to improve the level of BP prediction, using this non-destructive method, fruit could be examined periodically during the season to determine their elemental content. This is special important to supervise the effectiveness of Ca foliar applications. In addition, due to the advancement of XRF technologies, new instruments allow determining lighter elements such as N, B and Mg (micro-XRF) which could further improve the predictive power, considering their potential role on BP development.

5 Conclusions

The XRF reproducibility study in the calyx area of apple shows the potential of this tool for the development of ionomics at the individual fruit level. As a practical example, the easy calculation of K/Ca compared to the deconvolution process of K and Ca peaks would allow simplified monitoring of the K/Ca ratio under field conditions.

As discussed, given the infinite combinations of factors that influence the occurrence of BP (i.e., plant, climate, and management), it isn’t easy to venture which branch or part of it will produce fruit with more or less BP. Given the potential that classification methods would have for BP prediction, combining other databases that allow a more holistic understanding of the problem (e.g., plant reflectance) is relevant.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

CM, SR-B, MZ, RB, and GL contributed to the conception and design of the work. FG, DB, RC, RBe, JC and MB performed field experiment, acquisition, analysis, and interpretation of data for the work. CM, SR-B, MZ, RB, MB, and GL, collaborated to generate and validate the version to be published.

Funding

RB acknowledges support from Michigan State University (MSU) AgBioResearch, and the USDA National Institute of Food and Agriculture. In Chile, this work was funded by the National Agency of Research and Development (ANID; FONDEF ID18I10214) and FONDEQUIP EQM200239.

Acknowledgments

The authors sincerely thank DOLE Chile and Central Frutícola San Clemente for the supply of fruit and the permanent technical support.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2022.1033308/full#supplementary-material

References

Alarcón, A. L., Madrid, R., Romojaro, F., Molina, E. (1998). Calcium forms in leaves of muskmelon plants grown with different calcium compounds. J. Plant Nutr. 21, 1897–1912. doi: 10.1080/01904169809365531

CrossRef Full Text | Google Scholar

Al Shoffe, Y., Nock, J. F., Baugher, T. A., Marini, R. P., Watkins, C. B. (2020). Bitter pit and soft scald development during storage of unconditioned and conditioned ‘Honeycrisp’ apples in relation to mineral contents and harvest indices. Postharvest Biol. Technol. 160, 111044. doi: 10.1016/j.postharvbio.2019.111044

CrossRef Full Text | Google Scholar

Al Shoffe, Y., Nock, J. F., Watkins, C. B. (2014). Bitter pit and soft scald in ‘Honeycrisp’ during cold storage. HortScience 49, S203.

Google Scholar

Al Shoffe, Y., Nock, J. F., Zhang, Y., Zhu, L., Watkins, C. B. (2019). Comparisons of mineral and non-mineral prediction methods for bitter pit in ‘Honeycrisp’ apples. Sci. Hortic. 254, 116–123. doi: 10.1016/j.scienta.2019.04.073

CrossRef Full Text | Google Scholar

Amarante, C. V. T., Steffens, C. A., Ernani, P. R. (2010). Preharvest identification of bitter pit risk in “Gala” apples by fruit infiltration with magnesium and analysis of fruit contents of calcium and nitrogen. Rev. Bras. Frutic 32, 027–034. doi: 10.1590/S0100-29452010005000015

CrossRef Full Text | Google Scholar

Ballabio, D. (2015). A MATLAB toolbox for principal component analysis and unsupervised exploration of data structure. Chemom Intell. Lab. Syst. 149, 1–9. doi: 10.1016/j.chemolab.2015.10.003

CrossRef Full Text | Google Scholar

Ballabio, D., Consonni, V. (2013). Classification tools in chemistry. part 1: linear models. PLS-DA. Anal. Methods 5, 3790–3798. doi: 10.1039/C3AY40582F

CrossRef Full Text | Google Scholar

Ballabio, D., Grisoni, F., Todeschini, R. (2018). Multivariate comparison of classification performance measures. Chemom Intell. Lab. Syst. 174, 33–44. doi: 10.1016/j.chemolab.2017.12.004

CrossRef Full Text | Google Scholar

Bangerth, F. (1974). Investigations and reflections on the prediction of bitter pit incidence. Erwerbsobstbau 6, 169–172.

Google Scholar

Barker, M., Rayens, W. (2003). Partial least squares for discrimination. J. Chemom 17, 166–173. doi: 10.1002/cem.785

CrossRef Full Text | Google Scholar

Baugher, T. A., Marini, R., Schupp, J. R., Watkins, C. B. (2017). Prediction of bitter pit in ‘Honeycrisp’ apples and best management implications. HortScience 52, 1368–1374. doi: 10.21273/HORTSCI12266-17

CrossRef Full Text | Google Scholar

Bergmann, W. (1992). Nutritional disorders of plants: development, visual and analytical diagnosis (Jena; New York: Gustav Fischer).

Google Scholar

Biskup, S., Cosic, T., Pecina, M., Miljkovic, I. (2003). Relations of nutrient in apple leaves and fruits grown on acid soil and their influence on bitter pit occurrence. Pomol Croat 9, 25–35.

Google Scholar

Blume, H.-P., Brümer, G., Fleige, H., Horn, R., Kandeler, E., Kögel-Knaber, I., et al. (2016). Scheffer/Schachtschabel soil science (Berlin, Germany: Springer-Verlag).

Google Scholar

Bro, R., Smilde, A. K. (2014). Principal component analysis. Anal. Methods 6, 2812–2831. doi: 10.1039/C3AY41907J

CrossRef Full Text | Google Scholar

Burmeister, D. M., Dilley, D. R. (1991). Induction of bitter pit-like symptoms on apples by infiltration with Mg+2 is attenuated by Ca+2. Postharvest Biol. Technol. 1, 11–17. doi: 10.1016/0925-5214(91)90015-4

CrossRef Full Text | Google Scholar

Chamel, A. R., Bossy, J. P. (1981). Electron-microprobe analysis of apple fruit tissues affected with bitter pit. Sci. Hortic. 15, 155–163. doi: 10.1016/0304-4238(81)90103-5

CrossRef Full Text | Google Scholar

Conway, W., Sams, C., Hickey, K. (2002). Pre-and postharvest calcium treatment of apple fruit and its effect on quality. Acta Hortic. Sin. 594, 413–419. doi: 10.17660/ActaHortic.2002.594.53

CrossRef Full Text | Google Scholar

Espinosa-Zúñiga, C., Jarolmasjed, S., Sinha, R., Zhang, C., Kalcsits, L., Dhingra, A., et al. (2017). Spectrometric techniques for elemental profile analysis associated with bitter pit in apples. Postharvest Biol. Technol. 128, 121–129. doi: 10.1016/j.postharvbio.2017.02.009

CrossRef Full Text | Google Scholar

Fallahi, E. (2012). Influence of rootstock and irrigation methods on water use, mineral nutrition, growth, fruit yield, and quality in ‘Gala’ apple. HortTechnology 22, 731–737. doi: 10.21273/HORTTECH.22.6.731

CrossRef Full Text | Google Scholar

Fallahi, E., Conway, W., Hickey, K., Sams, C. (1997). The role of calcium and nitrogen in postharvest quality and disease resistance of apples. HortScience 32, 831–835. doi: 10.21273/HORTSCI.32.5.831

CrossRef Full Text | Google Scholar

Fallahi, E., Fallahi, B., Peryea, F. J., Neilsen, G. H., Neilsen, D. (2010). Effects of mineral nutrition on fruit quality and nutritional disorders in apples. Acta Hortic., 49–59. doi: 10.17660/actahortic.2010.868.3

CrossRef Full Text | Google Scholar

Fallahi, E., Fallahi, B., Valdés, C., Retamales, J. B., Tabatabaei, S. (2006). Prediction of apple fruit quality using preharvest mineral nutrients. Acta Hortic., 259–264. doi: 10.17660/ActaHortic.2006.721.35

CrossRef Full Text | Google Scholar

Fazio, G., Kviklys, D., Grusak, M. A., Robinson, T. (2013). Phenotypic diversity and QTL mapping of absorption and translocation of nutrients by apple rootstocks. Asp Appl. Biol. 119, 37–50.

Google Scholar

Ferguson, I. B., Triggs, C. M. (1990). Sampling factors affecting the use of mineral analysis of apple fruit for the prediction of bitter pit. NZJ Crop Hortic. Sci. 18, 147–152. doi: 10.1080/01140671.1990.10428086

CrossRef Full Text | Google Scholar

Ferguson, I. B., Watkins, C. B. (1992). Crop load affects mineral concentrations and incidence of bitter pit in `Cox’s orange pippin’ apple fruit. J. Am. Soc. Hortic. Sci. 117, 373–376. doi: 10.21273/JASHS.117.3.373

CrossRef Full Text | Google Scholar

Fernández, V., Díaz, A., Blanco, Á., Val, J. (2009). Surface application of calcium-containing gels to improve quality of late maturing peach cultivars. J. Sci. Food Agric. 89, 2323–2330. doi: 10.1002/jsfa.3727

CrossRef Full Text | Google Scholar

García-Sánchez, F., Galvez-Sola, L., Martínez-Nicolás, J., Muelas-Domingo, R., Nieves, M., Kyprianidis, K. (2017). “Using near-infrared spectroscopy in agricultural systems,” in Developments in near-infrared spectroscopy (London: IntechOpen), 97–127.

Google Scholar

Garrec, J. P. (1983). Modification de la composition minérale dans la zone de ‘suture molle’ (suture red spot) de pêches soumises à une pollution fluorée: Etude par microsonde electronique. Environ. pollut. A 30, 189–200. doi: 10.1016/0143-1471(83)90020-X

CrossRef Full Text | Google Scholar

Goode, J. E., Ingram, J. (1971). The effect of irrigation on the growth, cropping and nutrition of cox’s orange pippin apple trees. J. Hortic. Sci. 46, 195–208. doi: 10.1080/00221589.1971.11514398

CrossRef Full Text | Google Scholar

Hopfinger, J. A., Poovaiah, B. W. (1979). Calcium and magnesium gradients in apples with bitter pit. Commun. Soil Sci. Plant Anal. 10, 57–65. doi: 10.1080/00103627909366878

CrossRef Full Text | Google Scholar

Jarolmasjed, S., Espinoza, C. Z., Sankaran, S., Khot, L. R. (2016). Postharvest bitter pit detection and progression evaluation in ‘Honeycrisp’ apples using computed tomography images. Postharvest Biol. Technol. 118, 35–42. doi: 10.1016/j.postharvbio.2016.03.014

CrossRef Full Text | Google Scholar

Jarolmasjed, S., Khot, L. R., Sankaran, S. (2018). Hyperspectral imaging and spectrometry-derived spectral features for bitter pit detection in storage apples. Sensors 18, 1561. doi: 10.3390/s18051561

PubMed Abstract | CrossRef Full Text | Google Scholar

Jarolmasjed, S., Zúñiga Espinoza, C., Sankaran, S. (2017). Near infrared spectroscopy to predict bitter pit development in different varieties of apples. J. Food Meas Charact 11, 987–993. doi: 10.1007/s11694-017-9473-x

CrossRef Full Text | Google Scholar

Jemrić, T., Fruk, I., Fruk, M., Radman, S., Sinkovič, L., Fruk, G. (2016). Bitter pit in apples: pre- and postharvest factors: A review. Span J. Agric. Res. 14, 15. doi: 10.5424/sjar/2016144-8491

CrossRef Full Text | Google Scholar

Jungk, A., Claassen, N. (1997). Ion diffusion in the soil-root system. Adv. Agron. Vol. 61, 53–110. doi: 10.1016/S0065-2113(08)60662-8

CrossRef Full Text | Google Scholar

Kafle, G. K., Khot, L. R., Jarolmasjed, S., Yongsheng, S., Lewis, K. (2016). Robustness of near infrared spectroscopy based spectral features for non-destructive bitter pit detection in ‘Honeycrisp’ apples. Postharvest Biol. Technol. 120, 188–192. doi: 10.1016/j.postharvbio.2016.06.013

CrossRef Full Text | Google Scholar

Kalcsits, L. A. (2016). Non-destructive measurement of calcium and potassium in apple and pear using handheld X-ray fluorescence. Front. Plant Sci. 7. doi: 10.3389/fpls.2016.00442

CrossRef Full Text | Google Scholar

Kalcsits, L., Mattheis, J., Giordani, L., Reid, M., Mullin, K. (2019). Fruit canopy positioning affects fruit calcium and potassium concentrations, disorder incidence, and fruit quality for ‘Honeycrisp’ apple. Can. J. Plant Sci. 99, 761–771. doi: 10.1139/cjps-2019-0017

CrossRef Full Text | Google Scholar

Kim, M.-S., Ko, K.-C. (2004). Effects of forms and levels of nitrogen and levels of calcium on bitter pit incidence in ‘Fuji’ apples (Malus domestica borkh.). Hortic. Sci. Technol. 22, 200–205.

Google Scholar

Lang, A. (1990). Xylem, phloem and transpiration flows in developing apple fruits. J. Exp. Bot. 41, 645–651. doi: 10.1093/jxb/41.6.645

CrossRef Full Text | Google Scholar

Lee, L. C., Liong, C.-Y., Jemain, A. A. (2018). Partial least squares-discriminant analysis (PLS-DA) for classification of high-dimensional (HD) data: a review of contemporary practice strategies and knowledge gaps. Analyst 143, 3526–3539. doi: 10.1039/C8AN00599K

PubMed Abstract | CrossRef Full Text | Google Scholar

Le Grange, S. A., Theron, K. I., Jacobs, G. (1998). Influence of the number of calcium sprays on the distribution of fruit mineral concentration in an apple orchard. J. Hortic. Sci. Biotechnol. 73, 569–573. doi: 10.1080/14620316.1998.11511016

CrossRef Full Text | Google Scholar

López López, R., Arteaga Ramírez, R., Vázquez Peña, M. A., López Cruz, I., Sánchez Cohen, I. (2009). Índice de estrés hídrico como un indicador del momento de riego en cultivos agrícolas. Agric. Tec Mex 35, 97–111.

Google Scholar

Lötze, E., Joubert, J., Theron, K. I. (2008). Evaluating pre-harvest foliar calcium applications to increase fruit calcium and reduce bitter pit in ‘Golden delicious’ apples. Sci. Hortic. 116, 299–304. doi: 10.1016/j.scienta.2008.01.006

CrossRef Full Text | Google Scholar

Lötze, E., Theron, K. I. (2006). Existing pre-harvest predictions and models for bitter pit incidence. SA Fruit J. 5, 20–25.

Google Scholar

Marini, R. P., Baugher, T. A., Muehlbauer, M., Sherif, S., Crassweller, R., Schupp, J. R. (2020). Verification and modification of a model to predict bitter pit for ‘Honeycrisp’ apples. HortScience 55, 1882–1887. doi: 10.21273/HORTSCI15301-20

CrossRef Full Text | Google Scholar

Medrano, H., Bota, J., Cifre, J., Flexas, J., Ribas, M., Gulías, J. (2007). Eficiencia en el uso del agua por las plantas. Invest. Geogr. 43, 63–84. doi: 10.14198/INGEO2007.43.04

CrossRef Full Text | Google Scholar

Miqueloto, A., Amarante, C. V. T., Steffens, C. A., dos Santos, A., Mitcham, E. (2014). Relationship between xylem functionality, calcium content and the incidence of bitter pit in apple fruit. Sci. Hortic. 165, 319–323. doi: 10.1016/j.scienta.2013.11.029

CrossRef Full Text | Google Scholar

Mogollón, M. R., Contreras, C., de Freitas, S. T., Zoffoli, J. P. (2021). NIR spectral models for early detection of bitter pit in asymptomatic ‘Fuji’ apples. Sci. Hortic. 280, 109945. doi: 10.1016/j.scienta.2021.109945

CrossRef Full Text | Google Scholar

Mohr, W. P., Jamieson, G. M. (1984). The use of an electron microprobe to monitor the uptake of calcium and chloride in apples. Can. J. Plant Sci. 64, 349–354. doi: 10.4141/cjps84-049

CrossRef Full Text | Google Scholar

Nicolaï, B. M., Lötze, E., Peirs, A., Scheerlinck, N., Theron, K. I. (2006). Non-destructive measurement of bitter pit in apple fruit using NIR hyperspectral imaging. Postharvest Biol. Technol. 40, 1–6. doi: 10.1016/j.postharvbio.2005.12.006

CrossRef Full Text | Google Scholar

Oberdieck, J. G. C., Lucas, E. (1869). “Illustrierte monatshefte für obst-u,” in Weinbau (Ravensburg, Germany: Eugen Ulmer).

Google Scholar

Perring, M. A., Pearson, K. (1986). Incidence of bitter pit in relation to the calcium content of apples: Calcium distribution in the fruit. J. Sci. Food Agric. 37, 709–718. doi: 10.1002/jsfa.2740370802

CrossRef Full Text | Google Scholar

Peryea, F. J., Neilsen, G. H., Faubion, D. (2007). Start-timing for calcium chloride spray programs influences fruit calcium and bitter pit in ‘Braeburn’ and ‘Honeycrisp’ apples. J. Plant Nutr. 30, 1213–1227. doi: 10.1080/01904160701555077

CrossRef Full Text | Google Scholar

Prengaman, K. (2021). “An a-peeling approach for bitter pit prediction,” in Good fruit grower. (Washington State Fruit Commission, WA). Available at: https://www.goodfruit.com/an-a-peeling-approach-for-bitter-pit-prediction/.

Google Scholar

Retamales, J. B., León, L., Tomala, K. (2001). Methodological factors affecting the prediction of bitter pit through fruit infiltration with magnesium salts in the apple cv. ‘Braeburn.’. Acta Hortic. 564, 97–104. doi: 10.17660/ActaHortic.2001.564.10

CrossRef Full Text | Google Scholar

Retamales, J. B., Valdés, C., Dilley, D. R., León, L., Lepe, V. (2000). Bitter pit prediction in apples through mg infiltration. Acta Hortic. 512, 169–179. doi: 10.17660/ActaHortic.2000.512.17

CrossRef Full Text | Google Scholar

Saure, M. C. (2005). Calcium translocation to fleshy fruit: its mechanism and endogenous control. Sci. Hortic. 105, 65–89. doi: 10.1016/j.scienta.2004.10.003

CrossRef Full Text | Google Scholar

Seo, J. H., Heo, J. H., Choi, J. S., Ahn, Y. J. (2007). Crop load affects incidence of bitter pit and calcium contents in ‘Gamhong’ apple fruit. Hortic. Sci. Technol. 25, 110–113.

Google Scholar

Sharma, R. R., Pal, R. K., Singh, D., Singh, J., Dhiman, M. R., Rana, M. R. (2012). Relationships between storage disorders and fruit calcium contents, lipoxygenase activity, and rates of ethylene evolution and respiration in ‘Royal delicious’ apple (Malus × domestica borkh.). J. Hortic. Sci. Biotechnol. 87, 367–373. doi: 10.1080/14620316.2012.11512878

CrossRef Full Text | Google Scholar

Simons, R. K., Chu, M. C. (1980). Scanning electron microscopy and electron microprobe studies of bitter pit in apples. Acta Hortic. 92, 57–69. doi: 10.17660/ActaHortic.1980.92.7

CrossRef Full Text | Google Scholar

Sió, J., Boixadera Llobet, J., Rosera, J. (1999). Effect of orchard factors and mineral nutrition on bitter pit in ‘Golden delicious’ apples. Acta Hortic., 35, 331–334. doi: 10.17660/ActaHortic.1999.485.46

CrossRef Full Text | Google Scholar

Sió, J., Bosch-Serra, À.D., Rosera, J., Chiné, J., Boixadera Llobet, J. (2018). Características de los suelos y su fertilidad en relación con la incidencia de bitter pit en plantaciones de manzanas golden. ITEA 114, 206–222. doi: 10.12706/itea.2018.013

CrossRef Full Text | Google Scholar

Smock, R., Van Doren, A. (1937). The histology of bitter pit in apples. Proc. Amer Soc. Hort Sci., 176–179.

Google Scholar

Telias, A., Hoover, E., Rosen, C., Bedford, D., Cook, D. (2006). The effect of calcium sprays and fruit thinning on bitter pit incidence and calcium content in ‘Honeycrisp’ apple. J. Plant Nutr. 29, 1941–1957. doi: 10.1080/01904160600927492

CrossRef Full Text | Google Scholar

Terblanche, J. H., Gürgen, K. H., Hesebeck, I. (1980). An integrated approach to orchard nutrition and bitter pit control. Acta Hortic., 92, 71–82. doi: 10.17660/ActaHortic.1980.92.8

CrossRef Full Text | Google Scholar

Terblanche, J. H., Wooldridge, L. G., Hesebeck, I., Joubert, M. (1979). The redistribution and immobilisation of calcium in apple trees with special reference to bitter pit. Commun. Soil Sci. Plant Anal. 10, 195–215. doi: 10.1080/00103627909366888

CrossRef Full Text | Google Scholar

Torres, E., Alegre, S., Recasens, I., Asín, L., Lordan, J. (2021). Integral procedure to predict bitter pit in ‘Golden smoothee’ apples based on calcium content and symptom induction. Sci. Hortic. 277, 109829. doi: 10.1016/j.scienta.2020.109829

CrossRef Full Text | Google Scholar

Torres, E., Recasens, I., Lordan, J., Alegre, S. (2017). Combination of strategies to supply calcium and reduce bitter pit in ‘Golden delicious’ apples. Sci. Hortic. 217, 179–188. doi: 10.1016/j.scienta.2017.01.028

CrossRef Full Text | Google Scholar

Torres, E., Recasens, I., Peris, J. M., Alegre, S. (2015). Induction of symptoms pre-harvest using the ‘passive method’: An easy way to predict bitter pit. Postharvest Biol. Technol. 101, 66–72. doi: 10.1016/j.postharvbio.2014.11.002

CrossRef Full Text | Google Scholar

Tough, H. J., Park, D. G., Crutchley, K. J., Bartholomew, F. B., Graig, G. (1998). Effect of crop load on mineral status, maturity and quality of ‘Braeburn’ (Malus domestica borkh.) apple fruit. Acta Hortic. 464, 53–58. doi: 10.17660/ActaHortic.1998.464.4

CrossRef Full Text | Google Scholar

Val, J., Aznar, Y., Monge, E., Blanco, A. (1999). Un nuevo método de detección del bitter pit. Actas Horticul. 27, 188–192.

Google Scholar

Van Der Boon, J. (1980). Prediction and control of bitter pit in apples. i. prediction based on mineral leaf composition, cropping levels and summer temperatures. J. Hortic. Sci. 55, 307–312. doi: 10.1080/00221589.1980.11514939

CrossRef Full Text | Google Scholar

Volz, R. K., Alspach, P. A., Fletcher, D. J., Ferguson, I. B. (2006). Genetic variation in bitter pit and fruit calcium concentrations within a diverse apple germplasm collection. Euphytica 149, 1. doi: 10.1007/s10681-005-9000-8

CrossRef Full Text | Google Scholar

Volz, R. K., Ferguson, I. B. (1999). Flower thinning method affects mineral composition of `Braeburn’ and `Fiesta’ apple fruit. J. Hortic. Sci. Biotechnol. 74, 452–457. doi: 10.1080/14620316.1999.11511136

CrossRef Full Text | Google Scholar

Volz, R. K., Ferguson, I. B., Bowen, J. H., Watkins, C. B. (1993). Crop load effects on fruit mineral nutrition, maturity, fruiting and tree growth of ‘Cox’s orange pippin’ apple. J. Hortic. Sci. 68, 127–137. doi: 10.1080/00221589.1993.11516336

CrossRef Full Text | Google Scholar

von Bennewitz, E., Cooper, T., Losak, T., Hlusek, J., Carrasco-Benavides, M. (2015). Effects of decreasing levels of n amendments on organic ‘Granny smith’ apple trees. J. Soil Sci. Plant Nutr. 15, 979–990. doi: 10.4067/S0718-95162015005000068

CrossRef Full Text | Google Scholar

Weibel, F. P., Beyeler, C., Hauert, C. (2000). A strategy to prevent bitter pit incidence in organic apple growing by combining an early risk prediction on the basis of fruitlets analysis and specifically adapted soil management. Acta Hortic. 512, 181–188. doi: 10.17660/ActaHortic.2000.512.18

CrossRef Full Text | Google Scholar

Wills, R. B. H., Scott, K. J., Lyford, P. B., Smale, P. E. (1976). Prediction of bitter pit with calcium content of apple fruit. New Z. J. Agric. Res. 19, 513–519. doi: 10.1080/00288233.1976.10420983

CrossRef Full Text | Google Scholar

Wilms, Y. W. W., Basso, C. (1987). Soil amelioration for apple production in southern Brazil. Acta Hortic. 232, 193–203. doi: 10.17660/ActaHortic.1988.232.26

CrossRef Full Text | Google Scholar

Wold, S., Esbensen, K., Geladi, P. (1987). Principal component analysis. Chemom Intell. Lab. Syst. 2, 37–52. doi: 10.1016/0169-7439(87)80084-9

CrossRef Full Text | Google Scholar