Precision of estimating individual feed intake of grazing animals offered low, declining pasture availability

Abstract

Measurement of intake of individual grazing animals remains one of the fundamental challenges to improving efficiency of livestock production. The use of wireless sensor networks (WSN) shows potential for this purpose and requires benchmark data to underpin the necessary algorithm development. This study aimed to provide benchmark data and enable improved precision in estimating pasture intake when pasture availability is low and declining. Each of 10 Angus steers with a mean liveweight ± s.d. of 326 ± 46 kg was randomly allocated to an individual grazing plot. The plots comprised a monoculture of Italian ryegrass (Lolium multiflorum cv. Surge), with estimated initial pasture biomass availability ≤1100 kg DM/ha, provided at three levels of pasture availability (low, medium and high), achieved by varying plot sizes (0.2, 0.3 and 0.4 ha). Pasture intake was estimated using two pasture disappearance-based techniques (rising-plate meter and capacitance meter) using regression equations of daily pasture biomass estimates over an 11-day pasture intake period, and two chemical marker-based techniques (dosed n-alkanes and chromic oxide). Both pasture disappearance-based techniques showed high variability in estimating pasture biomass, with mean coefficients of variation between repeated-measurements of 28% for the capacitance meter and 44% for the plate meter, although daily biomass measurements over the duration of the study using the two devices were highly correlated (r = 0.82). Mean pasture intake estimates across all four techniques ranged from 3.4 to 10.7 kg DM/day. The estimates of pasture intake differed between techniques but not between biomass availability treatments. Mean of pasture intake estimates made using the plate meter were consistently higher than for the other three techniques. The correlation coefficients between the intake estimates determined using the pasture disappearance-based techniques, and between their rankings, were 0.61 and 0.58, respectively. Intake estimates obtained using pasture disappearance and the chemical marker methods were not correlated apart from between chromic oxide and the plate meter (r = 0.51). Further refinement of these techniques and more studies over a wider range of pasture conditions are needed. It is critical to understand the limits within which each of the pasture intake methodologies will produce reliable results that can then be used as benchmark data for the development of predictive algorithms using WSN.