Análise de trilha e espectroscopia no infravermelho próximo na cultura da canola

Abstract

The canola crop has been standing out among the main energy crops on the world stage and in Brazil its growth occurs especially in regions of high latitudes with a subtropical climate in the south of the country, but due to its high yield potential, there is a great interest on expanding its cultivation to the southeast region. In this sense, canola studies become crucial to expand its cultivation. Therefore, aiming to provide basis to future plant breeding programs as looking for more productives genotypes, the aim was to measure the effect of the association between related agronomic traits related to the yield of canola grains, grown at different sowing times regarding to the direct and indirect effects obtained from the path analysis. And also aiming speed evaluations up in plant breeding, the aim was to obtain an alternative method to the traditional chemistry, to quantify the oil content in grains, and setting a multivariate model which is able to predict the oil content in the grains through the analysis by spectroscopy in the near infrared region. The experiment was conducted in the field, in 2019 and a randomized-complete blocks design was used in split-plot system about the time, it was four spots (sowing season) and six subspots (canola hybrids) with four replicates. In each hybrid phenological observations were carried out to define the grain yield. Data were subjected to analysis of variance in the R program by the F test by 5 and 1% of probability. The oil content in the grains was determined by the traditional chemical method, and based on the near infrared spectral signature of grain samples, partial least squares regression (PLS-R) was established to estimate the oil content in canola grains. The sowing dates have influenced the yield components and also the oil content of the grains of all hybrids. The characters number of grains in five plants (0.6857) and height (0.4943) showed the highest estimates of positive correlation with grain yield, as well as the highest values of positive direct effect on yield, 0.2494 and 0.1595 respectively. In turn, the complete cycle (-0.7848), together with days in flowering (-0.4520), showed a significant negative correlation with yield. Thus, such characters in the canola crop deserve better attention when practicing selection in breeding programs in order to increase grain yield, especially the crop cycle that has a high direct negative effect on yield (-0.6157). The NIR technique associated with PLS-R was able to predict the oil content in the grains, resulting in good predictive models that can be used successfully in the analysis of the quality of samples after harvest and, thus, perform genotype selection faster. The best model presented an R2 of 0.86 and RMSE of 1.56 in the external validation, however, it is necessary to develop models with better variation in the oil content in the grains, expanding the number of samples to encompass all possible variations in the canola genotypes.