• Fecal NIRS for predicting percent leafy spurge in diets

      Walker, J. W.; Clark, D. H.; McCoy, S. D. (Society for Range Management, 1998-07-01)
      Research on diet selection is limited by inadequate techniques for determining botanical composition of diets. Our objective was to determine if near infrared reflectance spectroscopy (NIRS) of fecal material could be used to quantify the percentage leafy spurge (Euphorbia esula L.) in the diets of sheep (Ovis aries) and goats (Capra hircus). Fecal material representing diets of known percentage leafy spurge was obtained from feeding trials conducted in 1992 and 1994. In 1992, diets containing 87.5, 75, 60, 45, 30, and 15% leafy spurge were fed to 20 sheep and 20 goats. In 1994 10 sheep and 10 goats were fed alfalfa hay (Medicago sativa L.) at 0.5% of their body weight and ad libitum access to leafy spurge hay. Thus, the percent leafy spurge in the diet varied daily. Microhistological analysis was performed on fecal samples from the 1992 trial for comparison with NIRS predictions. Near infrared reflectance spectroscopy evaluations were performed with a scanning reflectance monochromator. Calibrations were done separately for sheep and goats. Samples were divided into calibration and validation sets. Data from the 1994 feeding trial were analyzed to determine the appropriate lag time between diet consumption and fecal spectral characteristics that provided the best prediction. The average of the diet composition 48 and 72 hours prior to the fecal sample provided the best predictions for the 1994 trial. The effect of spectral outliers on prediction accuracy was also evaluated. Spectral outliers were predicted with equal or better accuracy compared to samples that were spectrally similar to the ones from which calibration equations were derived. The NIRS predictions were more accurate than microhistological estimation of leafy spurge in the diet. The final calibration equation had coefficients of simple correlation for validation samples of 0.91 and 0.93 and standard errors of prediction of 4.6 and 4.8 for goats and sheep, respectively. The results of this study showed that NIRS of fecal material can be used to screen large numbers of animals for phenotypic differences in diet selection and for making treatment comparisons.