Métodos preditivos da porcentagem de sementes verdes e da longevidade de sementes de soja
Carregando...
Data
2021-02-23
Autores
Rezende, Brunna Rithielly
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Os investimentos em pesquisa levaram ao aumento exponencial da produtividade da soja nos últimos anos. O Brasil é o maior produtor mundial de soja, entretanto a cultura vem sofrendo com as mudanças climáticas, principalmente com altas temperaturas e baixa precipitações. Além de ser um fator que contribui com maior incidência de sementes verdes dentro do lote, o que afeta negativamente a qualidade fisiológica das mesmas. Os métodos usados para a separação de sementes verdes dentro do lote, são considerados demorados e ineficientes. Nesse sentido, o capítulo 1 propôs utilizar a tecnologia LIBS para classificar sementes verdes e amarelas dentro do lote. Assim, utilizou-se uma triplicata de 5 lotes de sementes de sojas que foram produzidas em condições de campo e, para o conjunto de dados, as linhas mais intensas do espectro foram selecionadas e aplicado 12 normalizações. Empregou-se Análise de Componente Principal (PCA), a qual foi viável de resposta e apresentou uma variabilidade >90.6%, mas não o bastante para indicar a diferença dos dois grupos (amarela x verde). A Análise Discriminante Linear (LDA) e a Quadrática (QDA) também foram empregadas mostrando que são eficientes na classificação, com precisão de 76.7% e 93.3% respectivamente. No outro capítulo, o objeto de estudo foi determinar a longevidade, período máximo de tempo em que a semente pode permanecer viável quando armazenada em condições ambientais favoráveis. Porém, a determinação da longevidade requer experimentos complexos e demorados e para sua predição é por meio de modelos matemáticos utilizando parâmetros como temperatura e umidade de armazenamento. Nesse sentido, o objetivo do capítulo 2 foi estimar a longevidade de sementes de soja por meio de variáveis de produção e de qualidade de sementes utilizando a Regressão Linear Múltipla (MLR) e de Mínimos Quadrados Parciais (PLS). Utilizou-se 47 genótipos de sementes de soja, os quais foram submetidos a testes laboratoriais a fim de obter informações sobre a qualidade de sementes e das variáveis de produção. Para o modelo PLS o coeficiente de determinação ajustado foi de 0,8404 utilizando todas as variáveis em estudo. Contudo, os métodos se mostraram eficientes na predição da longevidade de sementes de soja.
Research investments have led to the exponential increase in soy productivity in recent years. Brazil is the world's largest producer of soybean, however the culture has been suffering from climate change, mainly with high temperatures and low precipitation. In addition to being a factor that contributes more incidence of green seeds within the batch, which negatively affects the physiological quality of them. The methods used for separating green seeds within the batch are considered time consuming and inefficient. In this sense, Chapter 1 proposed to use LIBS technology to classify green and yellow seeds within the batch. Thus, a triplicate of 5 batches of seam seeds were produced under field conditions and, for the data set, the most intense lines of the spectrum were selected and applied 12 standardizations. Main component analysis (PCA) was employed, which was feasible for response and presented a variability> 90.6%, but not enough to indicate the difference of the two groups (yellow x green). Linear discriminant analysis (LDA) and quadratic (QDA) were also employed showing that they are efficient in classification, accurately from 76.7% and 93.3% respectively. In the other chapter, the study object was to determine longevity, maximum time period in which the seed can remain feasible when stored in favorable environmental conditions. However, the determination of longevity requires complex and time-consuming experiments and for their prediction is through mathematical models using parameters such as temperature and storage humidity. In this sense, the objective of Chapter 2 was to estimate soybean longevity through production and seed quality variables using multiple linear regression (MLR) and partial squares (PLS). 47 soybean seed genotypes were used, which were submitted to laboratory tests in order to obtain information on the quality of seeds and production variables. For the PLS model the adjusted determination coefficient was 0.8404 using all variables under study. However, the methods were efficient in predicting the longevity of soybeans.
Research investments have led to the exponential increase in soy productivity in recent years. Brazil is the world's largest producer of soybean, however the culture has been suffering from climate change, mainly with high temperatures and low precipitation. In addition to being a factor that contributes more incidence of green seeds within the batch, which negatively affects the physiological quality of them. The methods used for separating green seeds within the batch are considered time consuming and inefficient. In this sense, Chapter 1 proposed to use LIBS technology to classify green and yellow seeds within the batch. Thus, a triplicate of 5 batches of seam seeds were produced under field conditions and, for the data set, the most intense lines of the spectrum were selected and applied 12 standardizations. Main component analysis (PCA) was employed, which was feasible for response and presented a variability> 90.6%, but not enough to indicate the difference of the two groups (yellow x green). Linear discriminant analysis (LDA) and quadratic (QDA) were also employed showing that they are efficient in classification, accurately from 76.7% and 93.3% respectively. In the other chapter, the study object was to determine longevity, maximum time period in which the seed can remain feasible when stored in favorable environmental conditions. However, the determination of longevity requires complex and time-consuming experiments and for their prediction is through mathematical models using parameters such as temperature and storage humidity. In this sense, the objective of Chapter 2 was to estimate soybean longevity through production and seed quality variables using multiple linear regression (MLR) and partial squares (PLS). 47 soybean seed genotypes were used, which were submitted to laboratory tests in order to obtain information on the quality of seeds and production variables. For the PLS model the adjusted determination coefficient was 0.8404 using all variables under study. However, the methods were efficient in predicting the longevity of soybeans.
Descrição
Palavras-chave
Qualidade fisiológica, Classificação, LIBS, Longevidade, Predição, Modelos estatísticos, Physiological quality, Classification, Longevity, Prediction, Statistical models