Métodos estatísticos aplicados à seleção de clones de laranjeira 'Pêra'

• Ano de defesa: 2010
• Autor(a) principal: Souza, Emanuel Fernando Maia de
• Orientador(a): Siqueira, Dalmo Lopes de
• Banca de defesa: Bruckner, Claudio Horst , Silva, Fabyano Fonseca e
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Doutorado em Fitotecnia
• Departamento: Plantas daninhas, Alelopatia, Herbicidas e Resíduos; Fisiologia de culturas; Manejo pós-colheita de
• País: BR
• Palavras-chave em Português: Citrus sinensis; Laranjeira "Pêra"; Seleção; Repetibilidade; Medidas repetidas; Análise espacial
• Palavras-chave em Inglês: Citrus sinensis; Selection; Repeatability; Repeated measures; Spatial analysis
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::AGRONOMIA::FITOTECNIA::MELHORAMENTO VEGETAL
• Link de acesso: http://locus.ufv.br/handle/123456789/1139

Citrus industry has received prominence in the current economic and social scene for generating employment and income. Numerous studies have been developed in order to identify genotypes of superior citrus, enable an increase in genetic diversity of crops and reduce losses due to pests and diseases. Despite these advances, as for the orange trees, which are grown within the biggest planted area in Brazil, there are only a few works with the aim of selecting more productive and adapted clones to growing regions. Moreover, in competition assays of citrus growers, intended to obtain rootstocks or crowns, the determinations of the best genotypes use a large number of treatments, occupying large experimental areas for long periods, usually longer than eight annual crops, which makes these experiments lengthy and pricy. Therefore, the search for methods of analysis that can help increase efficiency and reliability in the data evaluation from these experiments is of supreme importance for the progress in the genetic improvement of citrus. This paper is aimed at identifying the minimum number of evaluations to be performed in competition assays of 'Pera' orange tree clones and propose the use of statistical methods to support the temporal or spatial dependence. At first, we determined the coefficient of repeatability and the phenotypic stabilization period to indicate the minimum number of annual crops with the intention of determining the superior clones. Then, models were analyzed in repeated measure analysis in order to assess temporal dependence as well as spatial dependence. The results show that 25 evaluations would be required for selection with 95% reliability. However, when considering the phenotypic stabilization period, i.e., the period in which there is greater expression of genetic characters that govern the production and less variability between successive crops, the selection could be accomplished by using the output of the first five harvests per year, seeing that in these crops there was more association with the total quantity of fruit production. By using statistics of the first five seasons, it would be possible to employ the repeated measurement model with the matrix of autoregressive heterogeneous covariance of first order, as this model features fit quality and selection efficiency at least equivalent or better than the method usually employed, i.e., the variance analysis with the average or total harvest per plant over the crops of interest. In view of spatial dependence control, the results indicate that the use of autoregressive separable first-order models, which were designed for selection experiments of orange tree clones, brought small, but significant gains. This fact may be due to the low values associated with the block variance and homogeneity of the experimental areas that were used. Therefore, the analysis, disregarding the block factor but with spatial autoregressive separable first-order adjustment, showed better fit quality among the models evaluated.