Crescimento de linho oleaginoso descrito por modelos de regressão não lineares

Abstract

The cultivation of linseed is an activity with high potential because it is a rustic plant, with low production costs and high demand in the domestic and foreign markets due to its nutritional and economic importance. However, it is little cultivated nationally due to the lack of studies on the cultivars and varieties used and the plant-atmosphere interactions. Thus, the objective of this study was to model the growth of linseed, using two varieties and two cultivars, cultivated in different agricultural years and sowing times, and adjusting nonlinear logistic and von Bertalanffy regression models, in order to indicate them as Statistical analysis tool to describe linseed growth. The data came from experiments carried out between 2014 and 2020, in the city of Curitibanos, Santa Catarina. The design was randomized blocks, with the treatments being the Dourada and Marrom varieties and the Aguará and Caburé cultivars, with four replications. Weekly evaluations were made of the number of leaves, plant height and number of secondary stems and, every two weeks, of total dry mass. The data were then organized into four collection methods: longitudinal, mean, random and cross-sectional, and subsequently tested in non-linear logistic and von Bertalanffy models. The best model was selected based on the value of the adjusted coefficient of determination, adjusted standard error, residual standard deviation, Akaike information criterion, Bayesian criterion and intrinsic and parametric non-linearity. In addition, the critical points of the model were obtained, namely the points of: maximum acceleration, inflection, maximum deceleration and asymptotic deceleration. The studied variables present a sigmoidal behavior, which allowed the adjustment of non-linear models, and among them, the logistic one was the most indicated, since it represents in a real way the estimates of the parameters and the critical points of the model, being an important way to evaluate growth variables of linseed. Among the data collection methods, there were better adjustments for the longitudinal, average and cross-sectional methods, the latter being considered an applicable alternative for the researcher in cases of need to reduce time, manpower or resources to conduct the experiment. From the logistic model, it was possible to infer about the growth of varieties and cultivars, in different years and sowing times, since the linseed cycle is directly related to the conditions of temperature, precipitation and sowing time. Thus, plant-atmosphere interactions are essential to understand the growth of agricultural crops, helping to choose management practices and ensuring high production rates. Although this work focuses on the linseed crop, the models are an analysis alternative for any agricultural crop.