Next Article in Journal
Effects of Humic Acids, Seaweed Extract and Equisetum arvense L. Extracts on Morphological, Histological and Physiological Parameters of the Ornamental Plant Ocimum basilicum Rokokó
Previous Article in Journal
Genome-Wide Identification of AGO, DCL, and RDR Genes and Their Expression Analysis in Response to Drought Stress in Peach
Previous Article in Special Issue
Variability Assessment of Banana Cultivars and Intercropping with Lemongrass Based on Fruit Quality Indicators
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Horticulturae
Volume 10
Issue 11
10.3390/horticulturae10111230
horticulturae-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Integrating Soil, Leaf, Fruitlet, and Fruit Nutrients, Along with Fruit Quality, to Predict Post-Storage Quality of Staccato Sweet Cherries

by
Mehdi Sharifi
1,*,
William Wolk
2,
Keyvan Asefpour Vakilian
3,
Hao Xu
1,
Stephanie Slamka
1 and
Karen Fong
1
1
Summerland Research and Development Centre, Agriculture and Agri-Food Canada, Summerland, BC V0H 1Z0, Canada
2
BC Tree Fruits Cooperative, Oliver, BC V0H 1T9, Canada
3
Department of Biosystems Engineering, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan 4913815739, Iran
*
Author to whom correspondence should be addressed.
Horticulturae 2024, 10(11), 1230; https://doi.org/10.3390/horticulturae10111230
Submission received: 22 October 2024 / Revised: 16 November 2024 / Accepted: 18 November 2024 / Published: 20 November 2024
(This article belongs to the Special Issue Orchard Management: Strategies for Yield and Quality)
Browse Figures
Versions Notes

Abstract

:
Predicting the post-storage quality of cherry fruits is crucial for determining their suitability for long-distance shipping or domestic distribution. This study aimed to forecast key quality attributes of Staccato sweet cherries after storage, simulating shipping conditions, by analyzing spring soil, leaf, fruitlet, and at-harvest data from thirty orchards in the Okanagan Valley, British Columbia, Canada, over two years. A support vector machine (SVM) was used to predict post-storage variables, with pre-harvest and at-harvest data selected by a genetic algorithm. The SVM accurately predicted soluble solids (R2 = 0.88), firmness (R2 = 0.83), and acidity (R2 = 0.79) after four weeks of storage, as well as visual disorders like slip skin and stem browning. Spring soil properties (Ca, Mg), leaf (N, Ca, Mg, Fe, Zn, B), and fruitlet data (N, Ca, Mg, B) were key predictors. Leaf Ca was vital for firmness and total soluble solids (TSS) prediction, while N in leaves and fruitlets influenced firmness, acidity, and disorders. Leaf Zn helped predict weight and acidity/TSS ratio, and Mg impacted fruit color. Pre-harvest leaf nutrition measured 3–4 weeks before harvest, proved most effective in predicting post-storage quality.

1. Introduction

The demand for high post-storage quality fruits is driven by long-distance refrigerated shipping [1]. Maintaining firmness and preventing skin disorders like decay, cracking, slip skin, pitting, and pebbling is essential for successful shipping [2]. While storage conditions matter, pre-harvest and at-harvest factors are crucial for post-storage outcomes [3]. Predicting post-storage quality helps producers optimize production and ensure long-term storage compatibility, which is vital for sweet cherry (Prunus avium), especially the Staccato variety, prone to quality deterioration during shipping. The Canadian cherry industry faces from three to ten million CAD in annual losses due to claims on 15–20% of exports (personal communication with Dr. Bill Wolk, Nov 2020). Despite these challenges, the impact of nutrient profiles during pre-harvest and at-harvest stages on post-storage quality remains unclear. Although Ross, Toivonen, Godfrey, and Fukumoto [2] identified links between foliar Ca, leaf size, and pitting or pebbling in cherries, strong correlations between pre-harvest, at-harvest variables, and post-storage quality are lacking.
Physicochemical analysis of fruitlets before harvest has proven effective in forecasting at-harvest and post-storage fruit quality. Studies show that fruitlet mass, calcium (Ca) levels, and the potassium (K)-to-Ca ratio, measured several weeks before harvest, can predict post-storage disorders in apples [3]. Early-season fruitlet Ca analysis has also been used to predict post-storage bitter pits in apples [4]. Nutritional content at harvest has been instrumental in forecasting parameters like firmness, TSS, and TA during storage. In sweet cherries, Ca improves fruit firmness, reduces pedicel shriveling, and decreases the incidence of fruit rots, highlighting the importance of nutrients in postharvest quality [5]. Conversely, nutritional imbalances at harvest can lead to defects such as softness, cracking, pitting, and stem browning in peaches [6]. Although research on various fruits, including mango [7], apple [8], kiwifruit [9], and sweet cherry [10], has explored the relationship between dry matter content and soluble solids at harvest, fewer studies have focused on pre-harvest and at-harvest nutrient concentrations and their impact on other post-storage quality traits in sweet cherries.
Tree leaf nutrient composition is also useful for predicting fruit quality. For example, Jivan, et al. [11] found that leaf nitrogen (N) significantly influenced fruit acidity (R2 = 0.690) in apples, while micronutrients had a greater effect on sugar accumulation (R2 = 0.809). Environmental factors such as temperature, solar radiation, and soil properties also affect fruit quality by influencing chemical composition [12]. However, there is still limited understanding of how these factors affect metabolic processes at the enzymatic and gene expression levels [13].
Several studies have investigated methods for predicting post-storage fruit quality, using approaches ranging from simple regression to advanced machine learning algorithms [4,14]. These studies have evaluated various input data, such as fruitlet and leaf characteristics and pre-harvest variables like soil composition [15]. Combining pre-harvest data with at-harvest fruit nutritional content has proven valuable for predicting post-storage quality [16]. Advanced modeling techniques, such as support vector machines (SVM), which can handle large datasets, may improve predictions of post-storage cherry quality [17]. These data-driven methods offer viable alternatives to conventional regression models, which are often limited by assumptions of linearity, normality, and variable independence [18]. Yan, et al. [19] successfully used artificial neural networks (ANN) to predict apple fruit quality based on soil pH and nutrient concentrations, achieving R2 values exceeding 0.65. Their analysis revealed that soil phosphorus (P), K, Ca, and magnesium (Mg) had the most significant impact on fruit quality.
While a few studies have analyzed both pre-harvest and at-harvest quality parameters for predicting post-storage cherry quality, nutrient-based prediction methods show promise. This study aims to determine whether post-storage quality attributes of Staccato sweet cherries can be predicted using spring soil properties, leaf, and fruitlet nutrient levels measured four weeks before harvest, and at-harvest fruit nutrients and quality. We hypothesize that nutrient concentrations in fruitlets, measured four weeks before harvest, can effectively predict post-storage quality and guide producers on whether to market their fruit domestically or internationally. To achieve this, we employed SVM as the primary predictor, along with a genetic algorithm (GA) for feature selection to identify the most relevant pre-harvest and at-harvest data for prediction tasks.

2. Material and Methods

2.1. Study Area

Thirty Staccato cherry orchard blocks representing a range of interior British Columbia growing conditions were selected in the Okanagan Valley (49°44′ N, 119°43′ W) (Figure 1). This region is characterized by cool winters (mean December–February temperature: −0.5 °C), warm summers (mean June–August temperature: 20.0 °C), and low annual precipitation (346 mm year−1) [20,21]. All orchard blocks were on Mazzard rootstock. Three adjacent trees of uniform training system (central leader), vigor, crop load, age, and similar trunk cross-sectional area were selected and labeled for the study in each orchard block in May 2021. These three trees were representative of the orchard and its growing conditions. All orchards were irrigated using microjet sprinklers. Trunk diameters at 0.3 m above ground level were measured to calculate the trunk cross-sectional area of selected trees in May 2021. Since the trees were large, and it was not feasible to harvest the entire tree to measure fruit load, a branch with a representative size and fruit load was selected. At harvest, the diameter of the branch was measured, and then all fruits on the branch were harvested. The fruit load was calculated by dividing the number of fruits by the branch’s cross-sectional surface area (CSA).

2.2. Sample Collection and Analysis

Soil, leaf, fruitlet, and fruit samples were collected from three pre-selected trees in each of the 30 orchards during both study seasons. For soil analysis, three composite soil samples, each consisting of three individual 5 cm diameter auger samples, were taken from each orchard site at a depth of 0–30 cm within the herbicide strip under the selected trees. This sampling occurred during the trunk measurement phase in May 2021.
Three to four weeks before the anticipated harvest, 60 fruitlet samples were collected from the mid-canopy of the three selected trees in each orchard block for nutrient analysis. The fruitlets were washed with distilled water, and their stems and pits were removed. The samples were then blended, and the resulting slurry was frozen for mineral analysis. At the same time, 30 recently fully expanded leaf samples (10 per tree) were collected from new lateral shoots at mid-branch height in each block. The leaves were dried, ground, and stored for later mineral analysis.
The harvest date for each orchard was determined by the respective grower based on the commercial maturity of the fruit. Fruits were uniformly collected from upright branches with a diameter of 1.5–3 inches, located 1–4 m above the ground, with sampling conducted from all cardinal directions (north, south, east, and west) at a 45–60 degree angle from the trunk. Three sets of fruit samples were collected at harvest from each orchard site: 25 healthy fruits per study tree, totaling 75 fruits per orchard; 50 randomly selected fruits per study tree, totaling 150 fruits per orchard; and 3 kg of fruit per study tree, totaling 9 kg per orchard.
From the 75 healthy fruits, subsamples were processed for analysis. Fifteen fruits (~100 g) were used for dry matter determination, where they were destemmed, pitted, weighed, sliced, and dried in an oven at 60 °C for 48 h, or until a consistent weight was achieved. Another 25 healthy fruits were assessed for firmness using a FirmTech instrument (FirmTech; Bioworks Inc., Wamego, KS, USA) and categorized into five maturity groups (very light, light, moderate, dark, and very dark) based on the Interprofessional Technical Centre for Fruits and Vegetables guidelines (CTIFL) (https://www.ctifl.fr/en/fruits, accessed on 15 June 2021). These fruits were also juiced for pH, TSS (°Brix), and TA (expressed as grams of tartaric acid per liter) analyses, with measurements conducted using a glass electrode, a Cole-Parmer digital refractometer, and automated titration with 0.1 mol L−1 NaOH to an endpoint pH of 8.1, respectively. The TA/TSS ratio was calculated as an indicator of the fruit flavor. A second set of 25 healthy fruits was washed, destemmed, pitted, blended, and stored as a slurry at −20 °C for nutrient analysis.
Sixty randomly selected fruit samples were weighed to determine the average fresh fruit weight. The fruits were then examined for the presence of specific disorders, including cracking, slip skin, stem browning, pitting, pebbling, and russet. Stem browning was classified as a “green stem” if it affected less than 10% of the stem. Disorders were identified and recorded based on the number of fruits exhibiting clear visual defects. Following the assessment, all samples were discarded. Ninety randomly selected fruit samples were processed for carposphere microbial assessments, which included quantifying total aerobic bacterial counts and yeast/mold counts on the fruit’s surface.
A sample of 8 ± 0.3 kg from the 9 kg harvest per orchard site was placed into a modified atmosphere bag (LifeSpan®), sealed, and packed into an 8.16 kg (18 lb) bulk cherry box following standard industry practices. Consistent handling practices were maintained during packaging. The samples were stored at 0.5 °C within two hours of harvest and evaluated for firmness and quality after 28 days.
After 28 days of cold storage at 0.5 °C, the fruit samples were removed and evaluated using the same methodology applied at harvest. Fifteen healthy fruits (~100 g) were used for post-storage dry matter determination. Twenty-five healthy fruits were assessed for post-storage firmness, pH, TSS (°Brix), TA, and the TA/TSS ratio. Another set of 25 healthy fruits was washed, destemmed, pitted, blended, and stored as a slurry at −20 °C for post-storage nutrient analysis. Additionally, 60 random fruit samples were weighed to determine the average fresh fruit weight and subsequently examined for post-storage disorders, including cracking, slip skin, stem browning, pitting, pebbling, and russet. Post-storage carposphere microbial assessments were conducted on 90 randomly selected fruits per orchard sample. All post-storage determinations followed the same protocols as those used for at-harvest measurements to ensure consistency and comparability.
Soil nutrient analysis included organic C, total N, nitrate (NO3−), P, K, Ca, Mg, S, Fe, Mn, Zn, copper (Cu), boron (B), sodium (Na), aluminum (Al), pH, cation exchange capacity (CEC), organic matter, and estimated N release (ENR). Organic C (after carbonate removal with 1.0 N HCl) and total N were measured using a LECO 628 CHN analyzer (LECO Corp., St. Joseph, MI, USA). The NO3− was extracted with 2 M KCl and analyzed colorimetrically using an Astoria-Pacific SFA (Astoria-Pacific Inc., Clackamas, OR). The P was extracted using Olsen-P, while K, Ca, and Mg were extracted with ammonium acetate (pH 7). Macronutrients and DTPA-extractable micronutrients were analyzed via ICP-OES (Spectro Analytical Instruments, Kleve, Germany), and B was extracted with hot water and analyzed by ICP-OES. The CEC was measured with ammonium acetate using ICP-OES. In total, 19 soil features were obtained. Finely ground leaf samples (<1 mm), and freeze-dried ground fruit and fruitlet slurries were analyzed for nutrients. Nitrogen concentration was measured using the LECO 628 CHN analyzer, while P, K, Ca, Mg, Fe, Mn, Zn, Cu, and B were measured with ICP-OES (ISOSpark Blue FMT26).

2.3. Statistics and Data Modeling

A comprehensive analysis was conducted on 62 distinct features, categorized as follows: 19 soil attributes, 10 leaf parameters, 11 fruitlet metrics, and 22 at-harvest fruit characteristics. These features were used to predict 13 post-storage fruit quality outcomes, including six physicochemical properties, one color metric, five fruit disorder traits, and the TA/TSS ratio. Fruit load data did not exhibit significant relationships with any other measured parameters and were therefore excluded from the report. Additionally, pre- and post-storage carposphere microbial assessments showed low coefficients of determination (R2 values below 0.2) with other variables. Consequently, the results of microbial analyses were not included in this paper.
The intelligent prediction system aimed to forecast post-storage fruit quality using pre-harvest or at-harvest data. A supervised learning method [22] generated a dataset that enabled the decision-making system to predict post-storage quality. Support Vector Machines (SVMs), a reliable technique widely applied in bioinformatics, image processing, and agriculture, were utilized.
Each sample was represented by feature vectors, and regression models were employed to establish relationships between input (RN) and output (R) spaces. Nonlinear regression was achieved by transforming data into a higher-dimensional space using ϕ: RN → RL. SVM regression solved the optimization problem [23] with the Gaussian kernel [24]. Given the large number of inputs, genetic algorithms (GA) were applied to optimize the model by iterating and generating robust solutions. The objective was to maximize the R2 value for predictions, with mutation rates, crossover percentages, and iterations set at 0.05, 0.50, and 500, respectively. The ReliefF algorithm [25] scored individual inputs to determine their relevance.
Model performance was evaluated using five-fold cross-validation, with features and targets normalized to avoid outliers. Due to the modest dataset size (60 samples), SVM was chosen, solving the optimization problem through a nonlinear transformation into a higher-dimensional feature space using the Gaussian kernel. Parameters were fine-tuned via GA to maximize the R2, where higher values indicated better prediction accuracy.
MATLAB 2018b was used to implement the machine learning tasks, including generating Taylor’s diagrams [26] for comparative evaluation. These diagrams assessed various input groups (soil, leaf, fruitlet, and at-harvest data) in predicting post-storage quality by comparing actual and predicted values, correlation coefficients, and root-mean-square distance (RMSD). Models with lower RMSD and higher correlation coefficients were considered more reliable. Taylor’s diagrams [26] highlighted the most effective data sources for predicting post-storage fruit quality, offering valuable insights for agricultural applications.

3. Results

Descriptive statistical analysis of variables measured across 30 Staccato cherry orchards in the Okanagan Valley, Canada, showed high variability (Tables S1–S5 in Supplementary Data), emphasizing the need for a robust modeling approach. Fruit load ranged from 1 to 43 fruits per branch CSA, with an average of 11.7 fruits per branch CSA. This variable showed no significant relationship with any of the soil, leaf, fruitlet, or fruit characteristics analyzed. Figure 2 presents the performance of SVM in predicting post-storage fruit attributes, using key features selected by GA from soil, leaf, fruitlet, and at-harvest data. The figure also shows predictions without at-harvest data, which are critical for timely decision-making. While at-harvest data improves accuracy, excluding it still yields useful predictions, though the largest performance drop was observed in fruit disorders.
In addition to the prediction of TA and TSS as two essential post-harvest quality indicators in sweet cherry fruits, the TA/TSS ratio—an important indicator of fruit flavor after storage—was also predicted using SVM. Additionally, the size of the fruit at harvest is significant, as larger fruits typically command higher market prices. However, since fruit size does not undergo substantial changes during storage, it was not considered in this study.
Although the study was limited to 60 samples, collected from 30 sweet cherry orchard blocks over two consecutive years, the total number of features was substantial. Machine learning methods generally struggle with high feature counts, as their learning process can be hindered by redundant features, many of which are unnecessary for predicting model output [27]. Table 1 identifies the effective features recognized by GA as predictors of post-storage fruit quality factors.
The R2 values in Figure 2 represent the meaning from five cross-validation runs. The SVM model effectively predicted key post-storage qualities of cherries, with the Gaussian kernel outperforming linear, polynomial, and sigmoid kernels, which were tested through trial and error. The highest accuracy was achieved for TSS (R2 = 0.88), followed by firmness (R2 = 0.83) and the TA/TSS ratio (R2 = 0.79). The lowest performance was observed for fruit weight (R2 = 0.41). SVM successfully predicted pitting (R2 = 0.73), while other disorders were predicted with R2 values ranging from 0.57 to 0.71 using at-harvest data. Without at-harvest data, predictions for disorders had lower R2 values, ranging from 0.32 (pebbling) to 0.45 (stem browning).
The effective features selected by GA and utilized by SVM to achieve the performances depicted in Figure 2 are detailed in Table 1. These features are sorted in descending order of their impact on the prediction task, starting with the highest-scoring feature and followed by those with lower scores. Since early prediction of fruit post-storage quality is crucial, the predictors measured before harvest are highlighted in bold in Table 1. The table shows the importance of pre-harvesting data. For example, although a high prediction performance (R2 = 0.88) was achieved for post-storage TSS in sweet cherry fruits using a combination of at-harvest fruit data along with soil and leaf variables, an R2 value of 0.71 was obtained using only leaf Fe and soil Ca.
Post-storage firmness was predicted accurately (R2 = 0.83) using one harvested fruit, one leaf, and two fruitlet attributes. The TA of fruits at harvest, along with leaf Ca and fruitlet N and B, were capable of predicting post-storage firmness. Using only leaf and fruitlet elemental composition as model inputs, it was possible to predict firmness with an R2 value of 0.69. Table 1 shows that leaf Zn and leaf Ca are among the most effective predictors for the TA/TSS ratio. Using only these two input variables, it was possible to predict the TA/TSS ratio with an R2 value of 0.70.
Generally, Table 1 shows that after the removal of harvested fruit properties from the predictions, leaf and fruitlet mineral content provide more useful data than soil composition in predicting output. Soil mineral content was the least effective factor in predicting the post-harvest quality of sweet cherry fruits. Some minerals in leaves and fruitlets, such as N, Ca, Fe, Zn, and B, were useful during predictions. The only useful soil characteristic was soil Mg, which was beneficial for predicting fruit color after storage. For fruit disorders, the highest score provided by the ReliefF algorithm (0.74) belonged to harvested fruit stem browning in predicting fruit slip skin. For these visual disorders, the results suggest an interconnected relationship: more stem browning during harvest led to more slip skin in fruits stored in cold temperatures, and vice versa.
To generate Taylor’s diagrams [26] in this study, each dataset was trained individually using SVM to predict output variables. A separate Taylor’s diagram [26] was created for each post-storage fruit quality factor, comparing actual data with SVM-predicted values (Figure 3). The diagrams highlight the contribution of each pre-harvest and at-harvest dataset. Blue points (at-harvest features) and, for some variables, green points (leaf features) closely match the actual data. For firmness, the minimal deviation between predicted and observed values shows that SVM can effectively predict post-storage firmness using only at-harvest data.
The predicted values based on soil properties showed the greatest deviation and lowest correlation with the actual data, with correlation coefficients generally below 0.6, indicating poor predictive performance. As a result, soil characteristics are not effective in forecasting fruit quality after storage. An exception is found in Figure 3i, which relates to the prediction of fruit pitting. Overall, while at-harvest fruit and leaf mineral data strongly correlate with post-storage quality, soil composition has a weak correlation with cherry fruit properties after storage.

4. Discussion

In this study, GA, an efficient feature selection algorithm, was used to identify the most effective features among all measured variables for training the prediction model. Selecting the most relevant features not only improves the performance of machine learning models [28], but also reduces experimental costs, as only three to four input features are needed to predict fruit post-storage quality, eliminating the need to measure all features analyzed in this study. Based on the results, while at-harvest fruit features played a significant role in predicting post-storage quality, several leaf and fruitlet properties also proved effective in forecasting post-storage characteristics. Our hypothesis that fruitlet nutrient concentrations alone, measured four weeks before harvest, could predict post-storage fruit quality was rejected. Among the pre-harvest factors, leaf data provided valuable insights into the post-storage condition of sweet cherry fruits.
The results demonstrated that the model most accurately predicted post-storage TSS, firmness, and the TA/TSS ratio. These parameters are critical indicators of fruit quality, as fruits that retain firmness, and flavor, and exhibit minimal visual defects after storage, are more marketable. Although other quality parameters were predicted with satisfactory accuracy, the model’s lower performance in predicting post-storage fruit weight may be due to the naturally high variability of this trait in the orchards studied.
N content in both leaves and fruitlets was a valuable predictor of post-storage characteristics such as firmness, TA, slip skin, and stem browning. Adequate N nutrition supports protein synthesis and cell wall integrity, contributing to improved fruit firmness [29]. The relationship between N content and fruit quality, including factors like color, carbohydrate availability, and maturity, varies across studies and growing conditions. For sweet cherries, Uçgun [30] found that increased N application improved fruit N content, N/K ratio, and firmness, while K fertilization had no significant effect on fruit quality but reduced fruit N content. Swarts, et al. [31] reported that higher N supply through fertigation increased fruit N concentration but decreased firmness at harvest and during storage. Additionally, N content influences fruit TA by affecting enzyme activity and organic acid metabolism. In oranges and tomatoes, higher N supply resulted in increased TA and organic acid content [32]. Higher TA levels in sweet cherries could also result from delayed fruit maturation due to high N supply [33].
Leaf Ca was a key predictor of post-storage fruit quality, particularly firmness, due to its role in strengthening cell walls and membranes, which reduces physiological disorders like cracking and enhances marketability [5,34]. Ca helps maintain fruit resilience and delays premature maturity by boosting cell turgor and membrane integrity [35]. Although the mechanisms of Ca uptake and distribution in fruits are not fully understood, they are influenced by fruit transpiration, which decreases as fruits mature [36]. Foliar Ca sprays, applied during pit hardening when cracking susceptibility increases, are crucial for preserving fruit quality [37]. The positive correlation between leaf Ca and fruit firmness likely stems from Ca’s role in pectic substance synthesis, reinforcing cell walls. Studies by Dar, et al. [38] and Quiroz, et al. [39] confirm that higher Ca levels improve firmness, while imbalances in K/Ca and N/Ca negatively affect it. Mid-season leaf Ca levels are reliable indicators of tree health, as Ca remains in leaf tissues without recycling under stress [40]. Our results showed that Ca concentrations in leaves and fruitlets also contributed to predicting post-storage fruit dry matter and visual disorders.
Soil and fruitlet Mg were effective in predicting post-storage fruit color. Mg is vital for plant growth, playing key roles in photosynthesis, carbohydrate transport, and the synthesis of nucleic acids and proteins [41]. Adequate Mg levels improve both crop yield and quality, while Mg deficiency reduces leaf chlorophyll and photosynthesis, negatively impacting these outcomes [42]. Research has also found that applying Mg at optimal levels significantly increases sucrose content during the fruit color-turning stage [43].
The levels of Fe and B in tree leaves were also effective variables in reliably predicting the post-storage quality of cherry fruits. Leaf Fe helped predict fruit pH and TSS, while leaf B content was useful in predicting fruit dry matter. Iron is essential for chlorophyll synthesis and is a component of various enzymes involved in photosynthesis and respiration. Adequate Fe levels in leaves ensure efficient photosynthesis, enhancing carbohydrate production and thus the TSS in fruits, like cherries. Ghesmati, et al. [44] demonstrated that foliar Fe application (as EDDHA, an organic chelated iron) significantly improved fruit quality traits, including TSS and total antioxidants. Boron, vital for sugar transport and cell wall development, influences fruit firmness [45]. In sweet cherries, B-fertilized trees had higher concentrations of TSS and anthocyanins than control plants [45]. Nagy, et al. [46] also noted that B fertilization affected B levels in leaf tissues, mostly after ripening.
Leaf Zn positively influenced the prediction of post-storage variables such as cherry weight, TA, and the TA/TSS ratio, likely due to its role in plant metabolism as an enzyme activator and auxin precursor [47]. Zn is essential for carbohydrate metabolism and protein synthesis, both of which are crucial for fruit development and size. Adequate Zn levels in leaf tissues support these processes, potentially leading to larger fruit weights [48]. Zn’s role in enzyme activation also impacts protein synthesis, carbohydrate metabolism, and seed production [49]. Additionally, Zn supports the synthesis of tryptophan, a precursor to indole acetic acid, which plays a key role in starch metabolism [47].
The development of visual disorders during cold storage can significantly reduce fruit quality, affecting both marketability and consumer appeal, especially in sweet cherries. This study found that pre-harvest data alone provided poor predictions for post-storage visual disorders while incorporating at-harvest fruit data significantly improved prediction accuracy. Many visual disorders manifest only during cold storage and are undetectable at harvest, which likely contributes to the limited effectiveness of machine learning algorithms in predicting these post-storage issues. The study acknowledges limitations stemming from the relatively small dataset, as well as the focus on a single sweet cherry variety and rootstock. We suggest that expanding data collection across a broader geographical range, incorporating multiple cherry varieties, diverse rootstocks, and varying climate conditions could substantially improve the accuracy and robustness of predictive outcomes.

5. Conclusions

This study highlights the effectiveness of machine learning algorithms, particularly SVM, in predicting post-storage sweet cherry quality by utilizing key pre-harvest and at-harvest data identified by a genetic algorithm. Most quality variables were predicted with R2 values between 0.6 and 0.9, demonstrating strong model accuracy. Pre-harvest data, especially mid-season leaf nutrients (N, Ca, Mg, Fe, Zn, and B), were reliable predictors of post-storage quality, while fruitlet nutrient concentrations alone were not sufficient. These findings emphasize the importance of early predictions for optimizing post-harvest management, enhancing profitability, and improving sustainability in cherry production. Elevated B levels in leaves and fruits, likely due to foliar applications, contrast with low soil B levels, raising questions about the physiological effects of low soil B in cherry production. Future studies should focus on refining leaf sampling strategies—timing, location, and leaf type—while expanding datasets could improve the accuracy of machine learning models, allowing for the exploration of more advanced algorithms. Advancing methodologies for in situ nutrient measurement, such as using portable X-ray fluorescence (XRF) technology, enhances real-time nutrient assessment and further streamlines the approach presented in this study.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/horticulturae10111230/s1. Table S1: Descriptive statistics of soil properties in the thirty studied Staccato sweet cherry orchards in Okanagan Valley, Canada. Table S2: Descriptive statistics of leaf nutrient concentrations in the thirty studied Staccato sweet cherry orchards in Okanagan Valley, Canada. Table S3: Descriptive statistics of fruitlets collected four weeks before harvest including nutrient concentrations and dry matter in the thirty studied Staccato sweet cherry orchards in Okanagan Valley, Canada. Table S4: Descriptive statistics of at-harvest fruit nutrient content, quality, and disorders in the thirty studied Staccato sweet cherry orchards in Okanagan Valley, Canada. Fruit color is not shown in this table due to its categorical nature. Table S5: The statistical variables of fruit post-storage properties measured in the study blocks. Fruit color is not shown in this table due to its categorical nature.

Author Contributions

M.S.: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing—original draft, Writing—review and editing. W.W.: Conceptualization, Funding acquisition, Methodology, Writing—review and editing. K.A.V.: Formal analysis, Validation, Visualization, Writing—original draft, Writing—review and editing. H.X.: Funding acquisition, Investigation, Methodology, Writing—review and editing. S.S.: Data curation, Methodology, Writing—review and editing. K.F.: Investigation, Methodology, Writing—review and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Agriculture and Agri-Food Canada, British Columbia Fruit Growers’ Association and British Columbia Cherry Association through Canadian Agricultural Partnership Program (Agriscience Projects—AGR-18383_ASP-160).

Data Availability Statement

Data will be made available on request.

Acknowledgments

The authors express their sincere gratitude to the British Columbia cherry growers who participated in this project. We extend special thanks to Erin Wallich of Summerland Varieties Corp. for her invaluable assistance in submitting our application to AAFC’s Canadian Agricultural Partnership program. We also acknowledge the support of BC Tree Fruits Cooperative and Jealous Fruits for providing cold storage facilities for the fruits. Additionally, we thank Masoumeh Bejaei for her consultation on statistical analysis and Gayle Krahn for her technical support in identifying orchards and coordinating sample collection from the Jealous Fruits-managed orchards.

Conflicts 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.

References

  1. Behdani, B.; Fan, Y.; Bloemhof, J.M. Cool chain and temperature-controlled transport: An overview of concepts, challenges, and technologies. Sustain. Food Supply Chain. 2019, 167–183. [Google Scholar]
  2. Ross, K.; Toivonen, P.; Godfrey, D.; Fukumoto, L. Pre-harvest conditions associated with Staccato sweet cherry fruit quality: Mineral status in leaves and fruitlets and orchard growing factors. Can. J. Plant Sci. 2020, 101, 262–267. [Google Scholar] [CrossRef]
  3. Nordey, T.; Davrieux, F.; Léchaudel, M. Predictions of fruit shelf life and quality after ripening: Are quality traits measured at harvest reliable indicators? Postharvest Biol. Technol. 2019, 153, 52–60. [Google Scholar] [CrossRef]
  4. Torres, E.; Recasens, I.; Àvila, G.; Lordan, J.; Alegre, S. Early stage fruit analysis to detect a high risk of bitter pit in ‘Golden Smoothee’. Sci. Hortic. 2017, 219, 98–106. [Google Scholar] [CrossRef]
  5. Winkler, A.; Knoche, M. Calcium and the physiology of sweet cherries: A. review. Sci. Hortic. 2019, 245, 107–115. [Google Scholar] [CrossRef]
  6. Meena, R.; Sarolia, D.; Reddy, S.; Kumar, R.; Ram, C.; Meena, A. Diagnosis and Management of Physiological Disorders in Arid Fruit Crops. In Dryland Horticulture; CRC Press: Boca Raton, FL, USA, 2021; pp. 327–337. [Google Scholar]
  7. Subedi, P.; Walsh, K.B.; Owens, G. Prediction of mango eating quality at harvest using short-wave near infrared spectrometry. Postharvest Biol. Technol. 2007, 43, 326–334. [Google Scholar] [CrossRef]
  8. Palmer, J.W.; Harker, F.R.; Tustin, D.S.; Johnston, J. Fruit dry matter concentration: A new quality metric for apples. J. Sci. Food Agric. 2010, 90, 2586–2594. [Google Scholar] [CrossRef]
  9. McGlone, V.A.; Jordan, R.B.; Seelye, R.; Martinsen, P.J. Comparing density and NIR methods for measurement of kiwifruit dry matter and soluble solids content. Postharvest Biol. Technol. 2002, 26, 191–198. [Google Scholar] [CrossRef]
  10. Toivonen, P.; Batista, A.; Lannard, B. Development of a predictive model for ‘Lapins’ sweet cherry dry matter content using a visible/near-infrared spectrometer and its potential application to other cultivars. Can. J. Plant Sci. 2017, 97, 1030–1035. [Google Scholar] [CrossRef]
  11. Jivan, C.; Sala, F. Relationship between tree nutritional status and apple quality. Hortic. Sci. 2014, 41, 1–9. [Google Scholar] [CrossRef]
  12. Lopresti, J.; Goodwin, I.; McGlasson, B.; Holford, P.; Golding, J. Variability in size and soluble solids concentration in peaches and nectarines. Hortic Rev. 2014, 42, 253–312. [Google Scholar]
  13. Minas, I.S.; Tanou, G.; Molassiotis, A. Environmental and orchard bases of peach fruit quality. Sci. Hortic. 2018, 235, 307–322. [Google Scholar] [CrossRef]
  14. Teh, S.L.; Coggins, J.L.; Kostick, S.A.; Evans, K.M. Location, year, and tree age impact NIR-based postharvest prediction of dry matter concentration for 58 apple accessions. Postharvest Biol. Technol. 2020, 166, 111125. [Google Scholar] [CrossRef]
  15. Doryanizadeh, M.; Ghasemnezhad, M.; Sabouri, A. Estimation of postharvest quality of “Red Delicious” apple fruits based on fruit nutrient elements composition. J. Agric. Sci. 2017, 9, 164–173. [Google Scholar] [CrossRef]
  16. Drouillard, A.; Léchaudel, M.; Génard, M.; Doizy, A.; Grechi, I. Variations in mango fruit quality in response to management factors on a pre-and post-harvest continuum. Exp. Agric. 2023, 59, e21. [Google Scholar] [CrossRef]
  17. Wang, X.; Bouzembrak, Y.; Lansink, A.O.; Van Der Fels-Klerx, H. Application of machine learning to the monitoring and prediction of food safety: A review. Compr. Rev. Food Sci. Food Saf. 2022, 21, 416–434. [Google Scholar] [CrossRef]
  18. Sarlaki, E.; Paghaleh, A.S.; Kianmehr, M.H.; Vakilian, K.A. Valorization of lignite wastes into humic acids: Process optimization, energy efficiency and structural features analysis. Renew. Energy 2021, 163, 105–122. [Google Scholar] [CrossRef]
  19. Yan, M.; Zeng, X.; Zhang, B.; Zhang, H.; Tan, D.; Cai, B.; Qu, S.; Wang, S. Prediction of Apple Fruit Quality by Soil Nutrient Content and Artificial Neural Network. Phyton 2023, 92, 193. [Google Scholar] [CrossRef]
  20. Environment and Climate Change Canada. Historical Climate Data. Available online: https://climate.weather.gc.ca/climate_normals/index_e.html#1981 (accessed on 28 May 2023).
  21. Houghton, E.; Bevandick, K.; Neilsen, D.; Hannam, K.; Nelson, L. Effects of postharvest deficit irrigation on sweet cherry (Prunus avium) in five Okanagan Valley, Canada, orchards: II. Phenology, cold hardiness, fruit yield, and quality. Can. J. Plant Sci. 2023, 103, 184–200. [Google Scholar] [CrossRef]
  22. Massah, J.; Asefpour Vakilian, K.; Torktaz, S. Supervised machine learning algorithms can predict penetration resistance in mineral-fertilized soils. Commun. Soil Sci. Plant Anal. 2019, 50, 2169–2177. [Google Scholar] [CrossRef]
  23. Massah, J.; Vakilian, K.A. An intelligent portable biosensor for fast and accurate nitrate determination using cyclic voltammetry. Biosyst. Eng. 2019, 177, 49–58. [Google Scholar] [CrossRef]
  24. Larada, J.I.; Pojas, G.J.; Ferrer, L.V.V. Postharvest classification of banana (Musa acuminata) using tier-based machine learning. Postharvest Biol. Technol. 2018, 145, 93–100. [Google Scholar]
  25. Robnik-Šikonja, M.; Kononenko, I. Theoretical and empirical analysis of ReliefF and RReliefF. Mach Learn. 2003, 53, 23–69. [Google Scholar] [CrossRef]
  26. Taylor, K.E. Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res. Atmos. 2001, 106, 7183–7192. [Google Scholar] [CrossRef]
  27. Javidan, S.M.; Banakar, A.; Vakilian, K.A.; Ampatzidis, Y. Diagnosis of grape leaf diseases using automatic K-means clustering and machine learning. Smart Agri. Technol. 2023, 3, 100081. [Google Scholar] [CrossRef]
  28. Javidan, S.M.; Banakar, A.; Vakilian, K.A.; Ampatzidis, Y. Tomato leaf diseases classification using image processing and weighted ensemble learning. Agron. J. 2024, 116, 1029–1049. [Google Scholar] [CrossRef]
  29. Torres-Olivar, V.; Villegas-Torres, O.G.; Domínguez-Patiño, M.L.; Sotelo-Nava, H.; Rodríguez-Martínez, A.; Melgoza-Alemán, R.M.; Valdez-Aguilar, L.A.; Alia-Tejacal, I. Role of nitrogen and nutrients in crop nutrition. J. Agric. Sci. Technol. B 2014, 4, 29. [Google Scholar]
  30. Uçgun, K. Effects of nitrogen and potassium fertilization on nutrient content and quality attributes of sweet cherry fruits. Not. Bot. Horti. Agrobo. 2019, 47, 114–118. [Google Scholar] [CrossRef]
  31. Swarts, N.; Mertes, E.; Close, D. Role of nitrogen fertigation in sweet cherry fruit quality and consumer perception of quality: At-and postharvest. Acta Hortic. 2017, 1161, 503–510. [Google Scholar] [CrossRef]
  32. Famiani, F.; Battistelli, A.; Moscatello, S.; Cruz-Castillo, J.G.; Walker, R.P. The organic acids that are accumulated in the flesh of fruits: Occurrence, metabolism and factors affecting their contents-a review. Rev. Chapingo Ser. Hortic. 2015, 21, 97–128. [Google Scholar] [CrossRef]
  33. Fallahi, E.; Righetti, T.L.; Proebsting, E.L. Pruning and nitrogen effects on elemental partitioning and fruit maturity in ‘Bing’sweet cherry. J. Plant Nutr. 1993, 16, 753–763. [Google Scholar] [CrossRef]
  34. Correia, S.; Queirós, F.; Ferreira, H.; Morais, M.C.; Afonso, S.; Silva, A.P.; Gonçalves, B. Foliar application of calcium and growth regulators modulate sweet cherry (Prunus avium L.) tree performance. Plants 2020, 9, 410. [Google Scholar] [CrossRef] [PubMed]
  35. Hepler, P.K. Calcium: A central regulator of plant growth and development. Plant Cell 2005, 17, 2142–2155. [Google Scholar] [CrossRef] [PubMed]
  36. Saure, M.C. Calcium translocation to fleshy fruit: Its mechanism and endogenous control. Sci. Hortic. 2005, 105, 65–89. [Google Scholar] [CrossRef]
  37. Kafle, G.K.; Khot, L.R.; Zhou, J.; Bahlol, H.Y.; Si, Y. Towards precision spray applications to prevent rain-induced sweet cherry cracking: Understanding calcium washout due to rain and fruit cracking susceptibility. Sci. Hortic. 2016, 203, 152–157. [Google Scholar] [CrossRef]
  38. Dar, M.; Wani, J.; Raina, S.; Bhat, M.; Malik, M. Relationship of leaf nutrient content with fruit yield and quality of pear. J. Environ. Biol. 2015, 36, 649. [Google Scholar]
  39. Quiroz, M.P.; Blanco, V.; Zoffoli, J.P.; Ayala, M. Study of Mineral Composition and Quality of Fruit Using Vascular Restrictions in Branches of Sweet Cherry. Plants 2023, 12, 1922. [Google Scholar] [CrossRef]
  40. Tang, R.-J.; Luan, S. Regulation of calcium and magnesium homeostasis in plants: From transporters to signaling network. Curr. Opin. Plant. Biol. 2017, 39, 97–105. [Google Scholar] [CrossRef]
  41. Chen, Z.C.; Peng, W.T.; Li, J.; Liao, H. Functional dissection and transport mechanism of magnesium in plants. Semin. Cell Dev. Biol. 2018, 74, 142–152. [Google Scholar] [CrossRef]
  42. Wang, Z.; Hassan, M.U.; Nadeem, F.; Wu, L.; Zhang, F.; Li, X. Magnesium fertilization improves crop yield in most production systems: A meta-analysis. Front Plant Sci. 2020, 10, 1727. [Google Scholar] [CrossRef]
  43. Liu, X.; Hu, C.; Liu, X.; Riaz, M.; Liu, Y.; Dong, Z.; Tan, Q.; Sun, X.; Wu, S.; Tan, Z. Effect of magnesium application on the fruit coloration and sugar accumulation of navel orange (Citrus sinensis Osb.). Sci. Hortic. 2022, 304, 111282. [Google Scholar] [CrossRef]
  44. Ghesmati, M.; Moradinezhad, F. Effect of foliar application of iron and zinc micronutrients on yield and quality properties of sour cherry fruit (Prunus cerasus L.). J. Hortic. Sci. 2019, 33, fa322. [Google Scholar]
  45. Wojcik, P.; Wojcik, M. Effect of boron fertilization on sweet cherry tree yield and fruit quality. J. Plant Nutr. 2006, 29, 1755–1766. [Google Scholar] [CrossRef]
  46. Nagy, P.T.; Thurzó, T.; Szabó, Z.; Nyéki, J. Impact of boron foliar fertilization on annual fluctuation of B in sweet cherry leaves and fruit quality. Int. J. Hortic. Sci. 2008, 14, 27–30. [Google Scholar] [CrossRef]
  47. Alloway, B. Zinc in Soils and Crop Nutrition. In International Zinc Association and International Fertilizer Association; International Zinc Association: Brussels, Belgium, 2008; p. 16. [Google Scholar]
  48. Marschner, H. Mineral Nutrition of Higher Plants, 3rd ed.; Academic Press: New York, NY, USA, 2012. [Google Scholar] [CrossRef]
  49. Swietlik, D. Zinc nutrition of fruit trees by foliar sprays. Acta. Hortic. 2002, 93, 123–129. [Google Scholar] [CrossRef]
Horticulturae 10 01230 g001
Figure 1. Location of experimental Staccato orchards in Okanagan Valley, British Columbia, Canada. Red symbols are the physical location of sweet cherry orchard used in the study.
Figure 1. Location of experimental Staccato orchards in Okanagan Valley, British Columbia, Canada. Red symbols are the physical location of sweet cherry orchard used in the study.
Horticulturae 10 01230 g001
Horticulturae 10 01230 g002
Figure 2. Performance of SVM in the prediction of post-storage quality of cherry fruits.
Figure 2. Performance of SVM in the prediction of post-storage quality of cherry fruits.
Horticulturae 10 01230 g002
Horticulturae 10 01230 g003aHorticulturae 10 01230 g003b
Figure 3. Taylor’s diagrams for the prediction of fruit postharvest attributes. Red-colored dots are the actual values of the attributes, while the other dots belong to the predicted values using features of fruitlets (orange), leaves (green), harvested fruits (blue), and soil (crimson). (a) dry matter, (b) weight, (c) firmness, (d) pH, (e) titratable acidity, (f) total soluble solids, (g) TA/TSS ratio, (h) slip skins, (i) stem browning, (j) pitting, (k) pebbling, and (l) russet.
Figure 3. Taylor’s diagrams for the prediction of fruit postharvest attributes. Red-colored dots are the actual values of the attributes, while the other dots belong to the predicted values using features of fruitlets (orange), leaves (green), harvested fruits (blue), and soil (crimson). (a) dry matter, (b) weight, (c) firmness, (d) pH, (e) titratable acidity, (f) total soluble solids, (g) TA/TSS ratio, (h) slip skins, (i) stem browning, (j) pitting, (k) pebbling, and (l) russet.
Horticulturae 10 01230 g003aHorticulturae 10 01230 g003b
Table 1. Predictive features for fruit post-storage quality and disorder parameters are listed, with each predictor accompanied by a value indicating its importance in predicting each attribute. Predictors highlighted in bold represent those measured before harvest.
Table 1. Predictive features for fruit post-storage quality and disorder parameters are listed, with each predictor accompanied by a value indicating its importance in predicting each attribute. Predictors highlighted in bold represent those measured before harvest.
Fruit Post-Storage AttributesEffective Predictors Among Soil, Leaf, Fruitlet, and At-Harvest Fruit Data
Dry matterHarvested fruit dry matter (0.44), Fruitlet Ca (0.22), Leaf B (0.21), Leaf Ca (0.21)
Cherry weightLeaf Zn (0.51), Harvested fruit Cu (0.28), Leaf Ca (0.24)
FirmnessHarvested fruit titratable acidity (0.23), Leaf Ca (0.19), Fruitlet N (0.17), Fruitlet B (0.15)
pHLeaf Fe (0.45), Leaf Ca (0.36)
Titratable acidityHarvested fruit Cu (0.45), Fruitlet N (0.39), Harvested fruit color (0.33), Harvested fruit Ca (0.19), Leaf Zn (0.19)
Total Soluble SolidsHarvested fruit dry matter (0.32), Leaf Fe (0.28), Soil Ca (0.26), Harvested fruit total soluble solids (0.14)
TA/TSSHarvested fruit total soluble solids (0.44), Leaf Zn (41), Leaf Ca (0.28)
ColorHarvested fruit color (0.21), Soil Mg (0.17), Fruitlet Mg (0.16)
Slip skinHarvested fruit stem browning (0.74), Harvested fruit slip skin (0.45), Fruitlet N (0.13)
Stem browning Leaf Ca (0.35), Harvested fruit stem browning (0.24), Leaf N (0.24), Harvested fruit Russet (0.19)
Pitting Harvested fruit slip skin (0.52), Harvested fruit stem browning (0.28), Leaf Ca (0.09)
Pebbling Harvested fruit pebbling (0.39), Harvested fruit weight (0.25), Leaf Ca (0.21), Harvested fruit B (0.20)
Russet Harvested fruit Ca (0.37), Harvested fruit stem browning (0.26), Harvested fruit russet (0.22), Fruitlet Ca (0.12)
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Sharifi, M.; Wolk, W.; Asefpour Vakilian, K.; Xu, H.; Slamka, S.; Fong, K. Integrating Soil, Leaf, Fruitlet, and Fruit Nutrients, Along with Fruit Quality, to Predict Post-Storage Quality of Staccato Sweet Cherries. Horticulturae 2024, 10, 1230. https://doi.org/10.3390/horticulturae10111230

AMA Style

Sharifi M, Wolk W, Asefpour Vakilian K, Xu H, Slamka S, Fong K. Integrating Soil, Leaf, Fruitlet, and Fruit Nutrients, Along with Fruit Quality, to Predict Post-Storage Quality of Staccato Sweet Cherries. Horticulturae. 2024; 10(11):1230. https://doi.org/10.3390/horticulturae10111230

Chicago/Turabian Style

Sharifi, Mehdi, William Wolk, Keyvan Asefpour Vakilian, Hao Xu, Stephanie Slamka, and Karen Fong. 2024. "Integrating Soil, Leaf, Fruitlet, and Fruit Nutrients, Along with Fruit Quality, to Predict Post-Storage Quality of Staccato Sweet Cherries" Horticulturae 10, no. 11: 1230. https://doi.org/10.3390/horticulturae10111230

APA Style

Sharifi, M., Wolk, W., Asefpour Vakilian, K., Xu, H., Slamka, S., & Fong, K. (2024). Integrating Soil, Leaf, Fruitlet, and Fruit Nutrients, Along with Fruit Quality, to Predict Post-Storage Quality of Staccato Sweet Cherries. Horticulturae, 10(11), 1230. https://doi.org/10.3390/horticulturae10111230

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop