Pasture management in Australia's dairy industry requires the manual shifiting of temporary electric fences to maintain pasture quality and growth. Virtual fencing presents an alternative to save time and labour costs. We used automated virtual fence (VF) collars to determine the variation in learning of the virtual fence stimuli, and evaluated the success of the technology to contain cows in a predetermined area of pasture. Twelve Holstein-Friesian non-lactating multiparous dairy cows were fitted with the collars, and a VF was used to restrict cows to two grazing allocations (G1 and G2) across six days. Cows received an audio tone (AT) when they approached the virtual fence, and a paired electrical pulse (EP) if they continued forward. The VF contained cows within predetermined areas for 99% of time, but cows spent the least time near the fence (p < 0.01). The number of stimuli reduced through time, demonstrating the ability of cows to learn the VF (p = 0.01). However, the mean number of EP per day ranged from 1 to 6.5 between individuals (p < 0.01). Therefore, successful containment may have a welfare cost for some individuals. Further work should focus on this individual variation, including measures of welfare.
Keywords: Virtual fence; associative learning; behavior; dairy; grazing; paddock usage; technology.