Non-destructive analysis of rapeseed quality by NIRS of small seed samples and single seeds

Abstract

Near infrared-reflectance spectroscopy (NIRS) is widely used as an efficient and non-destructive method for seed quality analysis in oilseed rape (Brassica napus L.). Normally, calibration equations are developed for seed samples sizes of 3-4 g. However, during the breeding process much less than 3 g of samples are frequently only available for analysis. The objective of the present study was to analyse whether calibration equations developed for standard sample sizes can be used to predict the oil, protein, moisture, glucosinolate and fatty acid content of sample sizes ranging from about 450 to 5 mg (single seeds). Special adapters were made for the standard ring cups to hold the smaller samples during NIRS scanning. Close Spearman rank correlations (r(S) = 0.82-1.00 ) showed that NIRS calibration equations developed for standard small ring cups can be used to predict the oil, protein, and glucosinolate content of smaller samples of 450-100 mg seeds. For oleic, linolenic and erucic acid, close correlations were only found for adapter sizes holding 450-250 mg seeds (r (S) = 0.79-0.98 ). Only moderate to very low correlations were found for the moisture content. The standard NIRS calibration equations could also be used to predict the oil content (R-2 = 0.79 ) and the protein content (R-2 = 0.72 ) of single seeds (ca. 5 mg), whereas this was not possible for the glucosinolate content (R-2 = 0.54 ). Therefore, single seed NIRS calibration equations were developed for alkenyl, indole and total glucosinolate content. The coefficients of determinations in cross validation ranged from 0.83 (alkenyl) to 0.86 (indole and total glucosinolates).