Virtual fencing without visual cues: Design, difficulties of implementation, and associated dairy cow behaviour

•A virtual fence (VF) system was implemented via wearable GPS technology.•The VF relied on GPS boundaries as opposed to traditional perimeter cables.•Dairy cows can be quickly trained to recognise visual cues for VF boundaries.•This association quickly deteriorated without continued visual reinforcement.•Cow comprehension of audio cues appears problematic without a visual guide. Intensive pasture-based farming systems rely on precise and frequent allocations of grass to animals. Virtual fence (VF) systems have been successfully used to contain animals within predefined boundaries. Accordingly, utilisation of a VF system to enhance automated allocation of correct forage areas to animals would represent a major advancement for grazing management strategies. Traditional VF systems rely on a perimeter cable to establish the boundary line, and this then needs to be deployed and physically moved to alter the parameters of the boundary. In our study, wearable GPS technology was used to implement a VF system without the need for such cabling. To accomplish this, we designed and developed a VF system comprised of a wearable collar with associated on-farm communication infrastructure. Moreover, we attempted to train dairy cows to associate an audio warning stimulus with boundary encroachment. Overall, the operating capacity of the cow-collar and the communications network were found to be robust. However, although dairy cows rapidly associated visual cues with VF boundary lines, and quickly developed a cue-consequence association between the audio warning and corrective stimulus, the number of boundary challenges made by cows increased upon removal of all visual cues. In addition, we observed a reduction in time spent grazing and ruminating during the training period, which suggested cows had become stressed within the designated inclusion zone. Nevertheless, our results are preliminary and further experimental work is required to truly assess best implementation protocols for virtual fencing without visual cues.