Submitted to: Transactions of the ASABE
Publication Type: Peer reviewed journal
Publication Acceptance Date: 7/1/2011
Publication Date: 7/1/2011
Publication URL: naldc.nal.usda.gov/download/53754/PDF
Citation: Armstrong, P.R., Tallada, J.G., Hurburgh, C., Hildebrand, D.F., Specht, J.E. 2011. Development of Single-Seed Near-Infrared Spectroscopic Predictions of Corn and Soybeans Constituents Using Bulk Teference Values and Mean Spectra. Transactions of the ASABE. 54(4):1529-1535. Interpretive Summary: Methods to measure the composition of single seeds of corn and soybeans would significantly enhance the ability of breeders to improve hybrids and varieties by selecting seeds that target breeder composition goals. Near-infrared reflectance spectroscopy (NIRS) is a method that has previously been used to measure single seed composition, such as protein, oil, and starch, but developing NIRS calibrations to do this generally requires the compositional measurement of single seeds. This cannot always be easily done and thus a method to calibrate for single seeds using bulk sample measurement was investigated as an alternative approach. NIRS calibrations using bulk sample methods were developed for corn protein, oil, starch, and kernel density; soybean protein, oil, and fiber. Results show that accurate seed measurement is possible although soybean fiber measurement is less accurate. These methods will allow NIRS calibrations to be developed more quickly and easily and should also expedite the development of commercial hybrids.
Technical Abstract: Near-Infrared reflectance spectroscopic prediction models were developed for common constituents of corn and soybeans using bulk reference values and mean spectra from single-seeds. The bulk reference model and a true single-seed model for soybean protein were compared to determine how well the bulk model performs in predicting single-seed protein. This provided a basis for evaluating bulk model performance for other constituents. Single-seed soybean oil was also predicted from the bulk model and sorted into bulk samples for reference analysis. Bulk model statistics indicated they should perform well for soybean protein and oil, but not as well for fiber; corn models should perform well for protein, oil, starch, and seed density. Bulk model predictions of single-seed soybean reference protein show that bulk models work reasonably well (SEP = 1.47%) but need bias correction (up to 4%) to be accurate. Bulk soybean oil predictions from bulk models and reference bulk values correlated well but were not accurate, indicating they can segregate into relative levels but do not provide absolute accuracy. In both cases, protein and oil, a slope and bias correction would increase accuracy significantly. Overall, the bulk models should be useful for selecting single-seeds in breeding programs targeting specific composition traits. It would be desirable to develop NIRS reference material which could be used to correct predictions and increase accuracy.