Summary
Several monitor herds were established on sown pastures at the CSIRO Lansdown Research Station near Townsville. In addition cattle in selected treatments of Queensland DPI grazing trials at Glentulloch near Injune, Galloway Plains near Calliope, Wambiana south of Charters Towers, and Rosebank near Longreach, were designated as monitor herds. Sampling requirements and measurements merely involved weighing the cattle at regular intervals of 4-6 weeks together with faecal sampling on weigh days (rectal grab samples) and mid-way between scheduled weigh days (sampling from fresh faecal pats). NIR spectra were obtained from dried, ground faecal samples. Reference values for growth rate were calculated from the cumulative liveweight gain curves for each monitor herd.
Calibration equations were developed in progression using a step-wise process of calibration, validation, expansion and recalibration. The first three calibration equations were developed using data from two monitor herds at Lansdown, firstly with the herd data kept separate and then by combining the two sets.
As data sets (faecal spectra and reference growth rates) from different sites/herds became available, each set was used first to test the predictive accuracy of the current, existing equation (validation test) and then to expand the calibration set for recalibration purposes. In this way the calibration set was built up to the current level of 629 samples comprising 32 herd-years (one herd-year represents the samples from a single monitor herd of one year’s duration) of information spanning the period June 1997 – May 2001 and 5 different sites.
None of the calibration equations developed to date has provided accurate predictions of growth rate when applied to samples unrelated to those in the calibration set (i.e. samples from different locations or different years) but there has been a clear trend for the predictive accuracy to improve as the calibration set expands. On the other hand, the actual calibration statistics (SEC, SECV, and RSQ) deteriorated with the expansion of the calibration set such that the SEC of the most recent equation stands at approximately 140 g/d. In the broad context this indicates that the probability of predicted growth rate being within 100 g/d or 200 g/d of actual growth rate is limited to approximately 60% and 85% respectively.
One of the difficulties with developing robust calibration equations for predicting growth rate is that of determining valid reference values from the regular weighing of cattle. Gut-fill and total body water account for a high proportion of the liveweight of cattle and changes in tissue weight (growth) that occur between successive weighings will generally be confounded to a greater or lesser degree by disproportionate changes in gut fill and body water. This means that “measured growth rates” used as reference values in developing calibration equations are likely to include a substantial error component and this will contribute to poor calibration statistics. Thus, the calibration equations are almost certainly somewhat better than the statistics indicate (Coates 2002). Similarly, validation tests are likely to exaggerate prediction errors.
Some other problem areas were identified. When cattle were in a compensatory growth phase, predicted growth rate was usually under-estimated. Conversely, when cattle were in a phase of rapid weight loss, predicted growth rate was often over-estimated (i.e. predicted weight loss was under-estimated). There was also an interaction between animal age/weight and the prediction errors. The errors were greater in older or heavier cattle, least in weaners. It is logical, too, that condition would also have an effect since the higher the condition score the greater will be the rate of weight loss when nutritional status declines. In a test case involving steers at Lansdown, predicted growth rates at points in time were often markedly different from measured growth rates, but calculated cumulative liveweight gain over an extended period was agreeably close to observed liveweight gain.
The weight of cattle relative to mature size has an effect on potential liveweight gain. Thus, annual liveweight gain of young cattle is usually greater than that of older, heavier cattle grazing the same pasture. It was satisfying to observe that faecal NIRS apparently coped with this age/weight/maturity effect on growth rate. Where weaner, yearling, and 2-year-old steers grazed the same pasture, predicted growth rates were usually highest for the weaners and lowest for the 2-year-old steers and the differences, when converted to cumulative liveweight gain, were consistent with the observed differences.
