Improved methodology to estimate intake of grazing animals is needed for better-inform management strategies as current techniques are limited. The objective of this study was to evaluate the accuracy of fecal near infrared reflectance spectroscopy (NIRS) to predict forage intake estimated using n-alkane markers in grazing animals. Fecal samples were collected from individual animals across 11 trials (N = 260) in which forage DMI was predicted using the alkane-ratio technique. For each trial, fecal samples were collected 2x daily for 5 consecutive d and composite samples subjected to NIRS analysis by a Foss NIRS 6500 scanning monochromator (Foss, Eden, Prairie, MN). Fecal spectra were used to develop equations to predict fecal alkane concentration (8 trials; N = 212) and n-alkane predicted DMI (11 trials; N = 260). For the prediction of fecal alkane concentrations, coefficients of determination for calibration (R
2c) and cross-validation (R
2cv) were 0.90 and 0.87 for fecal C
32 concentration, and 0.99 and 0.98 for fecal C
31 concentration. Calibration and cross-validation accuracies (R
2c and R
2cv) for the prediction of forage DMI estimated using the n-alkane method were 0.90 and 0.87, respectively. These results indicate the presence of strong associations between fecal NIRS spectra and fecal alkane concentrations, but do not provide information regarding the robustness of these equations, which is necessary for industry application. To evaluate the robustness of the equations in this study, independent-trial validation was performed. This type of validation was accomplished by removing a single trial from the data base and using the remaining 10 trials to develop the calibration equation to predict the independent trial. For this study, independent-trial validation results for the prediction of fecal alkane concentrations, and forage DMI estimated using the n-alkane method were poor (R
2v < 0.15). While cross-validation results indicate the potential of this technology to predict forage intake of grazing animals, the independent-trial validation results suggest that a larger data base will be needed to enhance robustness of predictive equations across diverse production systems.
|
|
|
|
|
|
|
|
|
Item |
N |
Range |
Mean ± SEL |
Calibration |
|
Validation |
R2c |
SEC |
R2cv |
SECV |
Fecal alkane concentration |
|
|
|
|
|
|
|
C31, mg/kg |
212 |
222-1564 |
986.4 ± 3.3 |
0.99 |
54.7 |
|
0.98 |
67.1 |
C32, mg/kg |
212 |
60-333 |
182.9 ± 2.5 |
0.90 |
20.6 |
|
0.87 |
23.2 |
N-alkane predicted DMI |
|
|
|
|
|
|
|
DMI, kg/d |
260 |
2.93-16.9 |
7.40 ± 0.16 |
0.90 |
0.74 |
|
0.87 |
0.84 |
SEL = standard laboratory error; SEC = standard error of calibration; SECV = standard error of cross validation
|