Use of NIRS faecal profiling to monitor the nutritional status of grazing ruminants

Donald Kneebone (2011). Use of NIRS faecal profiling to monitor the nutritional status of grazing ruminants PhD Thesis, School of Agriculture & Food Sciences, The University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s3433722_phd_finalthesis.pdf s3433722_phd_finalthesis.pdf Click to show the corresponding preview/stream application/pdf 2.15MB 13
Author Donald Kneebone
Thesis Title Use of NIRS faecal profiling to monitor the nutritional status of grazing ruminants
School, Centre or Institute School of Agriculture & Food Sciences
Institution The University of Queensland
Publication date 2011-09
Thesis type PhD Thesis
Supervisor Dr Gordon Dryden
Total pages 265
Total colour pages 19
Total black and white pages 246
Language eng
Subjects 07 Agricultural and Veterinary Sciences
Abstract/Summary Faecal near infrared spectroscopy (F.NIRS) profiling can be reliably used to measure many attributes of the diet of grazing ruminants. The research in this thesis describes the investigation of four areas in faecal profiling to extend the understanding and application of F.NIRS. These areas included a comparative study of the use of F.NIRS and conventional methods based on faecal chemistry to predict a number of dietary parameters; to broaden the existing understanding of whether F.NIRS could be extended to sheep fed diets comprised of various forages with or without supplements and how changes in feed chemistry were associated with faecal chemistry; to develop F.NIRS equations to assist with prediction of parameters associated with the nitrogen economy of the ruminant; and to investigate methods of validating prediction equations. In vivo data sets were obtained using sheep fed 25 different diets to provide 124 diet/faecal pairs. The diets consisted of five forages fed alone or with supplements of urea, molasses plus urea, cottonseed meal or sorghum grain. A calibration data set was developed using 99 samples that were selected randomly from the full set of samples, but ensuring that all diets were represented in the calibration set. Equations for a number of dietary parameters were developed using F.NIRS, as well as empirical models (F.CHEM) based on conventional chemical analysis. The remaining 25 samples were used to validate the F.CHEM and F.NIRS prediction equations. F.NIRS predictions for a number of dietary parameters, such as DM intake and DM digestibility with standard error of calibration (SEC) and standard error of cross validation (SECV) of 6.4 and 7.2 (g/kg0.75.d-1) and 0.031 and 0.035, respectively, were significantly better than predictions based on an F.CHEM model using faecal concentrations alone. Predictions using the empirical model were considered equivalent to F.NIRS only when that model was expanded to include both faecal concentrations and faecal excretion rates (F.CHEMe models). The F.CHEMe models had slightly lower standard errors of prediction (SEP) for OMI and digestible OMI, 6.3 (g/kg0.75.d-1) and 6.6 (g/kg0.75.d-1), respectively, compared to the F.NIRS models, while F.NIRS models had lower SEP for CPI (2.6 g/kg0.75.d-1) and OM digestibility (0.041). Good agreement between the F.CHEM and F.NIRS methods was evident (according to a 95% limits-of-agreement test), and both F.NIRS and F.CHEM predicted the reference values accurately and with small bias. Experiment 2(a) involved diets with 80% forage and 20% sorghum grain fed at six levels of intake, while in Experiment 2(b) diets were also fed at six levels of intake but had randomised forage and grain ratios. F.NIRS predictions for intake of DM, OM and CP were better for Experiment 2(a) than for Experiment 2(b). F.NIRS provided good predictions of intake with the proviso that changes in intake must be associated in a predictable way with changes in faecal chemistry. Samples from all three experiments were used to develop F.NIRS prediction equations for rumen microbial nitrogen synthesis. F.NIRS predictions for MN supply (g.d-1) were of moderate to good utility with validation R2 of 0.72 and SECV of 1.9 g.d -1 for Experiment 2(a), while calibration and validation R2 for Experiment 2(b) were 0.90 and 0.60, respectively. Predictions for allantoin excretion had a validation R2 of 0.87 and SECV of 1.2 mmol.d-1 for Experiment 2(a). Samples from Experiments 2(a) and 2(b) were used to develop additional F.NIRS equations for the prediction of metabolic faecal CP (MFCP) excretion with calibration and validation R2 for Experiment 2(a) being 0.99 and 0.86, and SEC and SECV of 0.2 and 0.3 g/kg0.75.d-1, respectively. MFCP excretion was estimated either as faecal protein soluble in neutral detergent, by regression of digestible protein on total dietary protein, or by F.NIRS equations developed using detergent-soluble protein as the reference value. No significant differences were evident between the detergent reference data and each of the prediction methods. Good predictions were obtained for a number of prediction sets based on single diet groups with or without supplements or prediction sets comprised of a number of forages supplemented with concentrates or energy supplements. One important result to come from this particular experiment was that to get a global equation calibration datasets must contain diverse samples and be validated using an external and unrelated data set. Two methods for preliminary examination of a set of samples were also examined. Principal component analysis and examination of average and difference spectra enabled spectral differences between the calibration and validation data sets to be identified rapidly. Both methods showed that different diet groups could be clearly identified from faecal spectra and provided an early indication of the likelihood of successfully developing prediction equations from a particular sample group. A further experiment involved the use of F.NIRS calibration equations developed specifically for cattle to predict dietary parameters for sheep. Data from Experiment 1 were used and the results obtained in this research did not indicate a uniform ability to predict dietary parameters between species. However it was evident that data sets, for example, based on forage-only diets could be used for prediction of dietary N concentration (SEP 0.35%), DM digestibility (SEP 0.046) and determination of faecal N (SEP 0.2%) for sheep. This research has shown that F.NIRS technology can be applied successfully to faecal profiling to extend the number and type of attributes that can be estimated, and to potentially strengthen the application of faecal NIRS technology to the nutritional management of grazing ruminants and native animals. Better management decisions have the potential to raise productivity and reduce overgrazing and the detrimental consequences that flow from the latter.
Keyword faecal NIR, supplements, intake, digestibility, purine derivatives, metabolic faecal nitrogen, prediction equations, internal and external validation
Additional Notes 101-103, 138-139, 196-202, 259-265

 
Citation counts: Google Scholar Search Google Scholar
Access Statistics: 168 Abstract Views, 13 File Downloads  -  Detailed Statistics
Created: Wed, 08 Feb 2012, 17:10:18 EST by Mr Donald Kneebone on behalf of Library - Information Access Service