Modelagem de percentuais de germinação de sementes de milho em função do tempo

• Ano de defesa: 2009
• Autor(a) principal: Gazola, Sebastião
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Programa de Pós-Graduação em Agronomia; UEM; Maringá, PR; Departamento de Agronomia
• Palavras-chave em Português: Milho; Germinação de sementes; Modelagem; Regressão não linear; Regressão probito; Inferência bayesiana.; Nonlinear regression; Seed germination; Probit analyses; Bayesian inference.; Ciências Agrárias; Agronomia
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/1155

In 2008, the national production of grains was about 58.7 million of tonnes for which the seed quality had the most effective contribution. However, the seed quality is reduced during warehouse storage and the prediction of seed lot performance under such conditions has been a challenge for seed technologist. Therefore, research programs must always investigate the seed lots performance in which the percentage of germination during the storage could properly be described by regression models. In this context, the germination performance of aged seeds of maize was estimated by the probit P(Y=y)=C+ (1-C).F(β0+β1.log(x)) as an alternative to the simplified equation in which Vp=Vі‾p.tg(β). Seeds from three seed lots of the hybrid maize OC 705 were aged at 43 °C for 24, 48, 72, 96, 120, 144, 168 and 192 h. The seeds were stored under warehouse conditions and the experiment was replicated three times. Coefficients of determination at most of 0.92 had no goodness of fit to describe the data by the simplified equation. On the other hand, the x² of the Pearson and log-likelihood tests indicated significant goodness of fit for the same data described by the probit model. Next, another analysis was carried out to fit the logistic model y(t)=C/(1+exp(B(t-M))) to the same data. The goodness of fit was evaluated by the parameter and intrinsic curvatures and bias of Box which were important in identifying the seed lot with the best fit. Finally, data from another experiment was analyzed to describe the germination performance of hybrid seeds of maize °C 705 and CD 5501 along sampling days and during two planting times by using the nonlinear model y(t)=A-B.exp(-Ct). First, the identity of the models indicated that in the seeding time 28 Oct. 1996 the performance from both hybrid seeds were described by the same equation. The date 28 Oct. 1996 was the best planting time and 57 d after anthesis was the best harvesting time when the percentage of seed germination was 96.1%. The methodology Bayesiana made possible the study of the germination curves and it allowed to recommend, for the two hybrid, the first harvesting time, as the viable for the sown. Still, for means of credibility intervals, it was possible the comparison of the adjusted equations for the combination of hybrid and harvesting times.