The Development of near Infrared Reflectance Spectroscopy Calibrations for Undried Grass Silage and Their Transfer to Another Instrument Using Multiple and Single Sample Standardisation

This study was undertaken to explore the possibility of developing accurate, robust near infrared (NIR) spectroscopy calibrations for undried grass silage (a very heterogeneous, high moisture forage) without costly or lengthy pre-treatment of the sample. This was achieved by scanning a large surface area of the sample using a coarse transport cell in conjunction with a Foss NIRSystems 6500 scanning monochromator, which moves the sample vertically past the light source. Equations were produced for alcohol corrected toluene dry matter, pH, N₂, neutral detergent fibre, acid detergent fibre, D-value (digestible organic matter in the dry matter) and potential metabolic intake with accuracy comparable to finely comminuted undried grass silage. The second objective of this study was to examine the possibility of transferring the undried grass silage calibrations developed on the 6500 spectrophotometer across NIR instruments and cell types. The cloning technique of Shenk was used to match spectrally the 6500 spectrophotometer and a Foss NIRSystems 5000 NIR spectrophotometer. Seventy undried grass silage samples, typical of the calibration set, were scanned through both instruments using a coarse transport cell and a natural product cell. Cloning was based on (a) 30 samples and (b) one central sample. Standardisation files were produced (using the ISI software) and applied to the spectra of the remaining 40 validation samples, scanned on the 5000 spectrophotometer (regarded as the slave instrument). These standardised spectra were predicted by the master equations developed on the 6500 instrument and the results compared to the corresponding master validation spectra (n = 40) scanned in the coarse transport cell and predicted by the master equations (regarded as the reference values). An alternative approach of standardising the master calibration set spectra, to look like they were scanned in a natural product cell and rerunning the calibrations, was also performed and the master natural product spectra predicted using the new natural product equations and compared to the reference values. In all instances, this cloning technique proved very successful, clearly indicating that undried grass silage calibrations can be transferred across these NIR instruments with little loss in accuracy of prediction.