[Study on developing calibration models of fatty acid composition in intact rapeseed by near infrared reflectance spectroscopy].

Four hundred ninety seven rapeseed samples, which feature multi-year, multi-loci and highly variant characteristics, were collected as a raw set. The NIR spectra were pretreated by scatter correction and mathematics treatments, and calibration models of fat acid composition of intact rapeseed were developed by using the algorithm method of modified partial least square (MPLS). Meanwhile, three types of sample cups with different capacity were used to screen the suitable calibration model for rapeseed quality breeding. The results showed that the calibration model of 8-gram-sample was the best, and the calibration determination coefficient was in the range of 0.74-0.98. The calibration effects of 3-gram-sample, which were similar to those of 0.6-gram-sample, were good with high determination coefficient (RSQ1, 1-VR) and low error (SEC, SECV). Therefore, the calibration set with multi-year and multi-loci samples can improve the accuracy and repeatability of NIRS models. The fat acids NIRS models of intact rapeseed developed could be introduced into breeding lines' selection, mutant screening and germplasm evaluation.