Multivariate statistical analysis applied to physical properties of soybean seeds cultivars on the post-harvest

. To consider the different characteristics of soybean seeds for designing and regulating the post-harvest equipment, we evaluated the similarities in the physical properties of soybean cultivars in this study. Two-hundred soybean seeds from 40 genetically modified cultivars were collected in packages to measure the physical properties of the seeds. First, principal component analysis was performed to verify the interrelationships between the variables and soybean cultivars. Next, a boxplot was constructed for each variable, considering the groups obtained after analyzing the main components. Finally, a scatterplot containing the Pearson's correlations between the variables was constructed. We identified two clusters of cultivars: C1 and C2. The unit-specific mass was the physical property that contributed the most to the formation of C1, whereas the other physical properties contributed to the formation of C2. Soybean cultivars comprising C1 were similar to each other only in unit specific mass, and the cultivars allocated to group C2 were similar according to all the other properties evaluated. These results can serve as a guideline for genotype selection for soybean genetic improvement to minimize variations in the physical characteristics of the seeds and obtain greater efficiency in the processing stages. Thus, the equipment manufacturing industry and seed processing units can implement projects and equipment adjustments to manage the post-harvest and seeding processes of soybean seeds efficiently.


Introduction
The use of precision agriculture techniques and the adoption of different soybean cultivars can increase soybean yield.Genetic breeding has enabled the development of a large number of soybean cultivars.Until May 2011, 823 soybean cultivars were listed in the National Register of Cultivars in Brazil out of which 502 cultivars were conventional and 321 were transgenic.The number of cultivars recorded in the National Register of Cultivars in January 2018 was 118% higher than that in 2011; out of the 1799 registered cultivars, 1234 cultivars were transgenic and 565 were conventional (Botelho, Granella, Botelho, & Garcia, 2015;Nikoobin, Mirdavardoost, Kashaninejad, & Soltani, 2009;Teixeira, Hampton, & Moot, 2020).
Post-harvest operations are affected by the variations in soybean genetic material, especially the physical variations in the seeds (Nikoobin, et al., 2009;Lima et al., 2021).Because of technological advancement, genetic improvement, and the production of different soybean seed cultivars, a variety of soybean seed lots with different physical characteristics are available.Hence, for greater operational yield in the post-harvest units, reduction of losses, and gains in quality, the development of flexible equipment (in terms of regulation and control) that consider the difference in the physical characteristics of seed cultivars is necessary (Araújo et al., 2020;Oliveira, Coradi, Alves, Teodoro, & Alvarez, 2021).
The composition and geometric shapes of seeds often interfere with the design, sizing, and regulation of machines used for handling, drying, storing, processing, and sowing (Jaques et al., 2022).The physical properties of a seed must be known for designing harvesting machines; ducts; discharge ramps; fans; sieves; and drying and aeration systems, determining the static capacity of the silos and conveyor belts, sizing the hoppers, separating, classifying, processing, handling, and storing (Nikoobin et al., 2009;Fernández-Fernández, Marcelo, Valenciano, López, & Pastrana, 2020).Studies conducted to evaluate the physical properties of the seeds of several crops have reported that uniformly sized seed lots ensures better postharvest operation performance, seed conservation, and quality (Payman, Ajdadi, & Bagheri, 2011;Mir, Bosco, & Sunooj, 2013).
For a more detailed analysis of the physical properties of a group of soybean seed cultivars multivariate statistical techniques must be applied (Antonucci et al., 2020).Using multivariate data analysis, the behavioral profile of a group exposed to the same phenomenon can be described based on all variables and their interactions (Mukasa et al., 2022).Thus, multiple measures can be analyzed simultaneously, which provide a deeper, more accurate, and more meticulous behavioral analysis.Recently, principal component analyses and Pearson correlations were adopted in some studies to investigate the drying and storage of seed cultivars and explain the results better (Coradi, Dubal, Bilhalva, Fontoura, & Teodoro, 2020a;Coradi et al., 2020b;Oliveira et al., 2021).
Thus, for considering the different characteristics of soybean-cultivar seeds and the different post-harvest operations and equipment, the aim of the study is to evaluate the similarity of the physical properties of seed cultivars and use it as a guiding parameter to decide the design and control of post-harvest equipment using multivariate statistical analysis.

Material and methods
In this study, we characterized the physical properties of six-hundred soybean seeds from each of 40 cultivars.The seeds were from heterogeneous lots produced and sown in the central-western region of Brazil.The soybean seeds from each sample harvested in 2019, 2020, and 2021 were stored in three different packages (top, middle, and bottom); each package contained 200 seeds.The shaken of the packages normally provides the accommodates smaller, medium, and larger seeds in different positions in the package.Thus, three sample points (top, middle, and bottom) were defined to characterize the seed lot better.In total, 136 lots of soybean seeds cultivars were evaluated.Seed size was determined by measuring the length, width, and thickness of the seed using a 0.01 mm resolution caliper, while the other physical properties were calculated using the equations listed in Table 1 (Mohsenin, 1989).
To evaluate the results, variance analysis was performed.The means were compared using the Scott-Knott test at 5% probability and we used the Sisvar 5.6 software to define the largest, intermediate-larger, intermediate-smaller, and smaller seeds.
Principal component analysis (PCA) was performed using standardized variables.A biplot was constructed with the first two principal components corresponding to due the accumulated variation in the components.In the biplot, two clusters were defined using the k-means algorithm, which groups observations into a cluster whose centroid is the closest to the observations until no significant variation is found in the minimum distance between each observation and the centroid.
A boxplot was constructed for each variable based on the groups obtained after analyzing the main components.The components were analyzed with the aid of the "qgraph", "ggfortify", and "ggplot2" packages in software R. Finally, for each cluster formed, a scatterplot containing the Pearson's correlations between the variables was constructed.A t-test was conducted to verify the significance of the correlations at 5% probability.
The projected area of seed cultivars averaged to 47.65 mm² and ranged from 59.71 in FLECHA IPRO to 36.58 mm² in cultivar NS 7447.The surface area of the seed cultivars averaged to 168.62 mm² and ranged from 214.27 in the FLECHA IPRO cultivar to 136.73 mm² in the TMG 1180GX RR cultivar.The surface-volume ratio averaged 0.773 and ranged from 0.834 in 2686 IPRO to 0.681 in FLECHA IPRO.The friction coefficient as a function of seed displacement averaged to 1.147 and ranged from 1.902 in cultivar 6139 RR to 0.650 in cultivar M6410.
The unit mass of seeds averaged to 0.186 g and ranged from 0.251 in cultivar FLECHA IPRO to 0.115 g in cultivar TMG 1180GX RR.The unit-specific mass of the seeds averaged to 1030.67 kg m -3 and varied from 1649.155 in cultivar KWS RK 6813 RR to 813.679 kg m -3 in cultivar 6139 RR.The apparent specific mass of the soybean cultivars averaged to 718.70 kg m -3 and ranged from 783.030 to 598.723 kg m -3 in cultivars FLECHA IPRO and 8473 RR, respectively.The resting angle averaged to 23.08° and ranged from 28.68° to 20.32° in cultivars KWS RK 6813 RR2 and 8473 RR, respectively.The average, maximum, and minimum porosities of the seeds were 39.20, 43.83 (in cultivar MAMBA RR), and 36% (in cultivar 8473 RR), respectively.
The seeds from cultivar UNIGEL 8473 RR Desafio (intermediate-larger seeds) exhibited considerable variation in apparent specific mass and porosity, while those from cultivar KWSRK 6813 RR (intermediatesmaller seeds) showed considerable variation in unit specific mass and angle of repose.The unit masses of cultivars FLECHA IPRO, NS 6823 RR, and TMG 1180GX RR were approximately the same and differed from that of all other cultivars.The coefficient of friction in cultivar AVANTSEED 96139 RR (intermediate-smaller seeds) was significantly different from that of the other cultivars.
The cultivar AVANTSEED SMANBA RR seeds (large seeds) had a higher porosity than others did.Seed lots need to be uniform for the optimal control of the post-harvest equipment; hence, three soybean cultivars exhibiting large size variations were separated to improve sowing quality and grain distribution.The length, width, and sphericity the seeds from cultivar UNIGEL NS 7447 differed from that of the other cultivar seeds, while the highest variations in thickness, volume, equivalent diameter, projected area, surface area, area ratio, and surface/volume, were found in the FLECHA IPRO cultivar seeds.The cultivar M7739 IPRO seeds showed the highest variations in circularity.
This finding was supported by the boxplot of physical properties, which showed a marked difference between the USMs of the clusters; cluster C1 exhibited a higher mean USM (Figure 2).The surface-to-volume ratio (SVR) and friction coefficient (FC) were the other properties for which C1 showed a higher mean than that of C2; however, they did not contribute to the formation of C1.
Acta Scientiarum.Agronomy, v. 46, e63664, 2024  Other physical properties contributed to the formation of C2.As can be observed from Figure 2, C2 had higher means for length (L), width (W), thickness (T), volume (V), circularity (Cr), sphericity (Sp), equivalent diameter (ED), and projected and surface areas (PA and SA, respectively).Hence, C2 cultivars had larger seeds than those allocated to C1.Based on the results provided by PCA and the boxplot of the physical properties, we inferred that the soybean cultivars comprising C2 were more similar to each other according to all the properties assessed, except for USM.USM was the property by which the cultivars allocated to the C1 group were the most similar.
The similarity results suggested that a certain group of soybean-cultivar seed lots had a higher homogeneity with respect to the physical properties.In this sense, the cultivars allocated to C2 were homogeneous in terms of volume, length, width, thickness, circularity, sphericity, equivalent diameter, projected area, surface area, SVR, friction coefficient, unit mass, apparent specific mass, rest angle, and porosity of seeds, whereas the C1-cultivars had a greater heterogeneity in terms of these properties.

Discussion
Owing to technological advances in genetic breeding and the production of different soybean seed cultivars, high variations exist in the physical characteristics of soybean seed lots.For higher operational yield in the post-harvest units, reduction of losses and gains in quality, the development of equipment that is flexible in terms of regulations and controls and considers groups of cultivar seeds similar in physical properties is necessary.However, in Brazil, the beneficiation and storage practices of soybean seeds still do not consider the heterogeneity of the seeds, which exist owing to the differences between cultivars.Given this reality, soybean cultivars similar in terms of physical properties need to be identified, which can help form batch groupings for post-harvest equipment adjustments.
The evaluation of the physical properties facilitated the identification of soybean seed cultivars with similar characteristics (C1 and C2) for obtaining more homogeneous seed lots for the pre-cleaning, drying, processing, and storage processes.Obtaining homogeneous seed batches with similar physical characteristics can contribute to the adjustment of each piece of equipment in each post-harvest process and the achievement of better seed quality.
Defining the characteristics of seed similarities enables equipment adjustment according to the seed lots received at the processing unit.Establishing quality control parameters in this manner is highly beneficial for the process.Homogenization of the seed batches handled during receiving, pre-cleaning, drying, processing, and storage minimizes the negative effects of operations on quality.This also allows a storekeeper to seize greater control of the variables in the operations, including flow, flow rate, seed mass, velocity, temperature, air pressure passing through the seed mass, and the dimensions, capacities, and adjustments of the equipments (Payman et al., 2011;Mir et al., 2013).
The identification of cultivars with similar physical seed properties contributes to increased efficiency in sizing, regulation, and maintenance of post-harvest equipment (Mohsenin, 1986, Mao et al., 2020).The clustering of cultivars according to physical seed characteristics allows control over the inclination and dimensions of the equipment during reception, transport, pre-cleaning, and processing; flow and air temperatures during drying; and aeration or cooling of seed mass during storage (Deshpande, Bal, & Ojha, 1993;Bande, Adam, Azni, & Jamarei, 2012;Mao et al., 2020).Homogenizing seed lots according to the properties of the cultivars improves the efficiency of drying systems and the quality of the seeds.It reduces the water content to levels that allow the conservation of water quality during storage (Deshpande et al., 1993;Bande et al., 2012;Jan, Panesar, & Singh, 2019).
Furthermore, knowing the physical properties of the lots from various cultivars facilitates the regulation of ventilation and cooling equipment used for reducing the temperature of the seed mass during storage and minimizing the changes in humidity.During storage, it helps in suppressing the respiratory processes that consumes nutritive reserves of the seeds and the changes linked to the metabolic dynamics that affect germination and vigor (Nori, Moot, & Mills, 2019;Pinheiro, Medeiros, Zavala-León, Dias, & Silva, 2020).
During batch reception, the seed-flow in a discharge hopper is gravity-based, and the inclination angle of the inner walls of the hopper is higher than the rest angle of the seeds.The sphericity, diameter, projected area, surface area, SVR, coefficient of friction, volume variability, and specific seed mass all affect the static capacity of the hopper.While handling and moving the seed mass using bucket elevators, conveyor belts, and threads, the mechanical damage inflicted to the seeds and the carrying capacity depend on adjustments, including those made according to the physical properties of the seeds.Such adjustments are often based on apparent specific mass, sphericity, projected area, surface area, SVR, coefficient of friction, and angle of repose of the seeds (Bande et al., 2012;Nori et al., 2019).
During a pre-cleaning operation, technicians separate impurities and foreign matter from the seed mass using air handling machines and sieves of different holes that are adjusted according to roundness, sphericity, equivalent diameter, and volume (length, width, and seed thickness) for smooth seed-mass flow and cleanliness (Moreano et al., 2018;Mao et al., 2020).
Drying alters the physical characteristics of the seeds.Thus, homogenization of seed lots by physical properties contributes to greater efficiency and uniformity in seed quality during the drying process.To make harvested seeds with high moisture content, usually in the range of 18-22%, more suitable for storage, they are subjected to artificial drying to reduce seed moisture to approximately 12% (w.b.).However, reducing the moisture content of seeds requires heat and mass transfer, which can substantially change the quality and physical properties of the seeds, depending on the method and drying conditions (Coradi et al., 2020b;Guilherme & Nicolin, 2020).
Drying causes changes in seed dimensions, especially a reduction in volume, which additionally alters the circularity, sphericity, equivalent diameter, volume, and specific mass of the seeds (Jung & Yoon, 2018).These changes result from the reduction in the tension inside the cells owing to the removal of water.Volumetric changes are the primary stimuli that cause changes in the physical properties of seeds, which determine the size and shape of the sieve holes used during the processing of agricultural products after harvest (Mao et al., 2020).The separation of soybean-cultivar seed lots into groups (C1 and C2) based on similar physical characteristics can reduce the effects of the drying processes, especially the effect of the drying-air temperature on seed quality.
The goal is to classify lots by size to improve quality and establish seed standards (Coradi et al., 2020b;Mao et al., 2020).Beneficiation is a fundamental operation in seed production programs and improves seed quality by providing conditions for use and meeting the minimum marketing standards established by legal norms (Oliveira et al., 2021).During processing, the seeds undergo several stages; however, not all lots follow the same sequence.This means that the operations in this process depend on the species, cultivars, and physical characteristics of the seeds (Jha & Kachru, 1998).The size and selection of the equipment for processing seeds are based on the physical characterization of the seeds.The flow during beneficiation is a time-consuming process and commonly results in mechanical injuries caused by physical agents during handling, which causes direct damage and render the seeds susceptible to contamination by highly deleterious pathogens (Ixtaina, Nolasco, & Tomas, 2008;Oksanen, 2018).
For seed processing, characterizing the seed lots (C1 and C2) according to the cultivar and its physical properties is essential, with emphasis on circularity, sphericity, volume, and specific mass.The quality of seeds is directly associated with the removal of inert material, seeds of weeds and other crops, and seeds of other cultivars, which depends on the proper selection of cleaning and separation equipment, the arrangement of the machines in the seed processing unit, and standardization of the physical seed properties (Sirisomboon, Kitchaiya, & Pholpho, 2007;Payman et al., 2011).A densimetric table is used to improve the physical quality of soybean seed lots by increasing the specific mass.It typically separates seeds based on the mass of one-thousand seeds (Atungulu & Olatunde, 2021).In a study conducted by Bakhtavar, Afzal, and Basra (2019), which aimed to evaluate the evolution of the physical characteristics (degree of humidity, percentage of impurity, apparent specific mass, and one-thousand seed mass) of the seeds of several soybean cultivars along the processing line, the authors concluded that processing reduces the impurities in many soybean seeds.The processing step is mainly conducted using a cleaning machine and completed using a densimetric table.
Atungulu and Olatunde (2021) reported that even with all the technologies available during the reception, pre-cleaning, and processing steps, qualitative and quantitative losses originate during storage when a seed mass is constantly subjected to external factors, such as temperature and relative humidity; chemical factors, such as the presence of oxygen; and biological factors, such as the development of bacteria and fungi (Khomari, Golshan-Doust, Seyed-Sharifi, & Davari, 2018).Therefore, minimizing such losses due to seed heterogeneity in the stages before the storage process is essential for increasing the efficiency and consequent profitability of the soybean seed production process (Homer, Patala, & Priedeman, 2015;Coradi et al., 2020b).

Conclusion
We obtained two groups of soybean seed cultivars with each group exhibiting similar physical characteristics.The USM was the physical property that contributed the most to the formation of C1, whereas the other physical properties contributed to the formation of C2.The soybean cultivars comprising C1 were similar to each other only with respect to USM.In contrast, the cultivars allocated to group C2 were more similar according to all the other properties evaluated.This is the first study in which cultivars were clustered based on the main physical properties of soybean seeds and the cultivars with the highest uniformity for the properties assessed were identified.We recommend these results for the development and adjustment of equipment for seed processing.Furthermore, these results can be used as guidelines to selection the genotypes for the genetic improvement of soybean to minimize variations in the physical characteristics of the seeds and obtain greater efficiency in soybean-seed processing stages.

Table 1 .
Equations for determination of physical properties of soybean seeds.

Table 2 .
Evaluation of physical properties of different soybean seed cultivars.

Table 3 .
Evaluation of physical properties of different soybean seed cultivars (Continued Table 2)