|Reeves Iii, James|
Submitted to: Journal of the Science of Food and Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/16/1999
Publication Date: N/A
Interpretive Summary: Mid-infrared spectroscopy is a non-destructive assay method using a form of light which can be used to determine the composition and quality of materials. Samples of tall fescue (a forage or grass used as animal feed) from three growing years, ground by two methods, and oven and freeze dried were analyzed chemically and by mid-infrared spectroscopy for protein, fiber and digestible dry matter. Results indicated that the same principles which govern the development of near-infrared calibrations (another method using light, which has been widely used to determine the composition of materials such as fescue) apply to the development of calibrations using mid-infrared spectra. In both of these methods spectra (graphical representation of how the light interacts with the sample) are used to develop equations which can then be used to determine the composition of other samples. Overall, results indicated that spectra collected at a higher resolution (more detailed spectra) produced better results, that oven and freeze dried samples yield equally good results, but, unlike previous near-infrared results, that how the samples are ground can affect the results, with more finely and consistently ground samples giving better results. Reducing the size of the calibration set (number of samples used to develop the equations) resulted in a decrease in calibration accuracy similar to that found with near-infrared calibrations. In conclusion, the results presented in this paper show that mid-infrared spectra can be used for the determination of the composition of forages and behave similar to those developed using near-infrared spectra.
Technical Abstract: The objective of this study was to determine the feasibility of using mid- infrared diffuse reflectance spectroscopy to determine the composition (fiber and crude protein) of diversely treated forages. Study samples (325) were randomly selected from a maximum of 2808 possible sample/treatment combinations, generated using eighteen feedstuffs, 26 treatments (chemical and physical), and six treatment levels. Neutral detergent fiber and crude protein were determined on all samples and calculated on an organic matter basis. Samples were scanned (256 scans each) from 4000 to 400 cm-1 at 4 and 16 cm-1 resolutions on a FTIR spectrometer equipped with a sample transport device, and on a scanning monochromator from 9,100 to 4,000 cm-1 (rotating sample cup, 64 scans, 10 nm resolution). Stepwise calibration results, using all samples in the calibration set showed calibration results for samples scanned at 4 cm-1 in the mid-infrared to be virtually identical to those achieved on the monochromator: R2 and SEC for neutral detergent fiber and crude protein of: .876, 4.10 and .928, 1.51, and .887, 3.92 and .926 and 1.54, for the near- and mid-infrared respectively. Partial least squares regression analysis with a single rotation sample were about the same for crude protein, but significantly poorer for neutral detergent fiber with R2 and RMSD of .642, 7.79 and .709, 6.73, for the near- and mid-infrared, respectively. In conclusion, while mid-infrared spectroscopy was able to determine the composition of a diverse set of treated samples, it was not a significant improvement over the near-infrared. Finally, partial least squares did not perform as well for neutral detergent fiber as stepwise regression.