High-troughtput phenotyping of physiological and yield related traits in soybean using hyperespectral sensor

• Ano de defesa: 2021
• Autor(a) principal: Paula, Ramon Gonçalves de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Soja - Melhoramento genético; Folhas; Reflectância; Sensoriamento remoto; Previsão de rendimento; Melhoramento Vegetal
• Link de acesso: https://locus.ufv.br//handle/123456789/28749

Increasing yield still one of the main challenges in soybean breeding programs due to its complexity as a result of the great genetic variability and genotype x environment interaction. Whereas great advantage has been achieved in selecting top yielded cultivars in last decades, breeders have frequently face unfavorable environment situation for selection, such as drought stress caused by lack of rainfall at critical phases of the crop. Better understanding of the phenotype might provide additional information about the genotype and help in a more accurate selection. Assessing physiological, biochemistry, and morphology information of the genotyping is becoming necessary to broken down complex traits and achieve better results in plant breeding. Traditional methods to assess these traits is not feasible in large scale breeding company, so the use of fast, noninvasive and accurate tools such as remote sensing is becoming usual. This study tested the use of leaf hyperspectral reflectance (350 – 2500 nm) in 20 soybean cultivars as a high- throughput phenotyping approach to: (1) estimate physiological trait under drought condition and selected cultivars based on these traits; and (2) predicted yield-related traits and to verify the availability on the selection procedure based on those traits. Partial least square (PLS) regression models accurately estimated seven out of eight physiological traits when the measurements were taken 20 and 28 days after drought imposition. Built PLS discriminant analysis model allowed to select the top five cultivars, having 100% of coincide when comparing to the FAI-BLUP selection index regarding all physiological traits simultaneously. Reliable PLS regression models prediction were also found for six out of ten yield-related traits, including number of seeds per pound and 100 grain weight. Good association was found between the selected wavelengths and FAI-BLUP indexes values, indicating that the model could be used as a selection approach. The results showed in this study indicate the great potential in using hyperspectral reflectance as a feasible, nondestructive, and accurately method to phenotyping physiological and yield-related traits as well as to screening superior genotypes. Keywords: Leaf hyperspectral reflectance. High-throughput phenotyping. Physiological traits. Yield components. Drought condition. Soybean.