An efficient method for screening seed colour in Ethiopian mustard using visible reflectance spectroscopy and multivariate analysis

Summary Seed colour is considered as an important breeding objective in Ethiopian mustard and other Brassica species, but its measurement is usually subjective rather than objective. For this reason we studied the potential of visible reflectance spectroscopy (VRS) combined with factorial analysis and multivariate calibration to predict seed colour in seeds of Ethiopian mustard. A total of 8331 samples were screened in the range 400–2500 nm with a near infra-red reflectance spectroscopy (NIRS) spectrophotometer. A common factor analysis was performed using 1148 samples and spectral data from 400 to 700 nm. It was found that four common factors explained 99.9% of the variance. Three of these factors were associated with colour characteristics of the seeds. Factor 1 separated yellow, yellow-brown and brown seeded samples, factor 3 separated brown from reddish-brown samples and factor 4 separated complete yellow from greenish-yellow samples. Calibration equations were developed for each factor, which allowed seed colour in those samples not included in factor analysis to be predicted. This study demonstrated that seed colour prediction can be incorporated into NIRS routine analysis with instruments that incorporate the visible spectral region.