Development of a nirs calibration for important quality parameters in maize

Maize is an important crop having a high potential of use in human and animal feeding as well as in industrial areas. The grain?s quality traits are often as important as the yield since they determine the area of end use. Easier, faster and cheaper methods are needed in place of standard laboratory procedures both in industry to determine the quality aspects of the products and in breeding programs to evaluate the genetic stocks. In this respect, NIRs technique has been emerged as an important alternative. In this study, development of calibration models were attempted based on partial least squares (PLS) regression model to estimate carbohydrate, oil, protein, and moisture ratios in maize grain using FT-NIR instrument. NIR spectra were measured on intact and ground seed samples from 150 maize genotypes, and these data were correlated in order to develop calibration models. Also, estimation accuracies of two different commonly used protein analyses methods in NIR spectra based models were compared. Results showed that, in all calibration models, ground seed samples yielded (for Kjeldahl r=0.84; for elemental analysis r=0.92; for oil ratio r=0.73; for carbohydrate ratio r=0.70; for moisture ratio r=0.77) better estimations than intact seed samples (for Kjeldahl r=0.65; for elemental analysis r=0.84; for oil ratio r=0.63; for carbohydrate ratio r=0.65; for moisture ratio r=0.54). The most accurate estimations were obtained with the models for protein ratio. To estimate protein ratio, regression model developed based on elemental analysis was more successful than that based on Kjeldahl method. The result suggested that protein ratio estimations using FT-NIR spectroscopy were in acceptable accuracy levels, while, different methods needed to be developed to increase the power of calibration models for the other traits.