Temporary confinement during parturition and early postpartum may provide an intermediary step preceding loose housing that offers improvement in sow and piglet welfare. Three experiments were conducted to investigate the implications of replacing farrowing crates (FCs) with an alternative housing system from 3 days postpartum until weaning. In each experiment sows farrowed in FCs and were randomly allocated at day 3 of lactation to either a FC or a pen with increased floor space (lactation pen (LP)) until weaning. In experiment 1, piglet growth and sow and piglet skin injuries were recorded for 32 sows and 128 focal piglets in these litters. Behaviour around nursing and piglet behavioural time budgets were also recorded for 24 of these litters (96 focal piglets for time budgets). In experiment 2, measures of skin injury and behavioural time budgets were conducted on 28 sows and 112 focal piglets. The behavioural response of sows to piglet vocalisation (maternal responsiveness test (MRT)) was also assessed. In experiment 3, piglet mortality from day 3 of lactation until weaning was recorded in 672 litters over 12 months. While housing did not affect piglet weight gain in experiment 1, or piglet skin injuries in experiments 1 or 2, sows in both experiments sustained more injuries in LP than FC (experiment 1, 2.9 v. 1.4; experiment 2, 2.5 v. 0.8 lesions/sow; P<0.05). Sow–piglet interactions were more frequent in LP than FC at days 11 and 18 postpartum in both experiment 1 (day 11, 1.4% v. 1.2%; day 18, 1.7% v. 1.0% of observations; P=0.05) and 2 (day 11, 1.0% v. 0.3%; and at day 18 were 1.0% v. 0.6% of observations; P<0.01), and LP sows were more responsive in the MRT in experiment 2 (2 v. 0 median number of tests in which sows react, P<0.01). In experiment 1 piglets played more (0.7% v. 0.3% of observations, P=0.05) and manipulated others less (0.3% v. 0.7% of observations, P=0.04) in LP, but more piglets missed nursing bouts (0.2 v. 0.1 piglets/bout, P<0.01) compared with FC. There was no effect of housing on piglet mortality from day 3 of lactation until weaning in experiment 3 (0.63 and 0.64 deaths/litter for LP and FC, respectively, P>0.05). Thus, housing sows and litters in LP from day 3 of lactation minimises piglet mortality while improving maternal behaviour in sows and social behaviour in piglets.

Virtual fencing technology uses a neckband-mounted device to deliver an audio cue when an animal nears a virtual boundary that is set via a global positioning system, followed by an electrical stimulus if it crosses the boundary. The flexibility offered by this technology could revolutionize grazing management on dairy farms, but its application and effects on lactating dairy cattle have not been assessed. This experiment reports on the effects of an electric or a virtual front-fence on dairy cow behavior and welfare. Two temporally separated treatments were applied to a herd of 30 multiparous cows. Cows were provided an estimated 14 to 15 kg of dry matter/cow of fresh pasture in a new paddock every 24 h. From d 1 to 10 cows were grazed using a conventional electric front-fence (control treatment) and from d 14 to 23 they were grazed using a virtual front-fence (eShepherd, Agersens Pty Ltd.). Cows were trained to the technology from d 11 to 13. The milk production and live weight of individual cows were recorded daily. Cortisol concentrations were obtained from milk samples collected from individual cows on 3 d during each of the control and the virtual fence grazing periods, plus the first day of training. From d 6 of the experiment, 6 focal cows were fitted with a RumiWatch (Itin + Hoch GmbH) noseband sensor to monitor grazing and ruminating time, and 8 focal cows were fitted with an IceTag (IceRobotics Ltd.) sensors to monitor activity. Milk production, live weight, and the time cows spent standing and lying did not differ between the electric and virtual fence periods. Milk cortisol concentrations, activity, and the times spent ruminating and grazing were comparable between the electric and early virtual fence periods (i.e., d 1-3 with a virtual fence). However, at d 4 to 6 with a virtual fence, activity (steps taken and motion index) and time spent grazing were lower, and time spent ruminating was greater, compared with an electric fence. Further, least significant difference tests suggest milk cortisol concentrations were higher at d 5 with a virtual fence than at d 8 with an electric fence and d 1 with a virtual fence. We conclude there is no evidence of behavioral and welfare effects of virtual fencing on dairy cows in the days immediately following implementation of the technology in a simple intensive grazing regimen, but a longer study is required to fully elucidate effects beyond this period.

Virtual fencing is promoted as the next advancement for rotational grazing systems. This experiment compared the capacity of conventional temporary electric versus virtual fencing to contain a herd of 30 lactating dairy cows within the boundaries of their daily pasture allocation (inclusion zone). Cows were moved each day to a new rectangular paddock that was divided crosswise into an inclusion and exclusion zone by a single linear electric (first 10 d) or virtual (second 10 d) front-fence. A 3-d virtual fence training period separated the 2 treatments. Virtual fences were imposed using a pre-commercial prototype of the eShepherd virtual fencing system (Agersens Pty Ltd.). Neckband-mounted devices replaced the visual cue of an electric fence with benign audio cues, which if ignored were accompanied by an aversive electrical stimulus. Cows learned to respond to the audio cues to avoid receiving electrical stimuli, with the daily ratio of electrical to audio signals for individual cows averaging (± standard deviation) 0.18 ± 0.27 over the 10 d of virtual fence deployment. Unlike the electric fence, the virtual fence did not fully eliminate cow entry into the exclusion zone, but individual cows were generally contained within the inclusion zone ≥99% of the time. Pasture depletion within the inclusion zone reduced the efficacy of the virtual fence in preventing cows from entering the exclusion zone, but the magnitude of this effect was insignificant in practical terms (i.e., increased time spent in the exclusion zone by ≤28 s/h per cow). This highlights the potential for virtual fences to control grazing dairy cow movement even when pasture availability is limited (i.e., 1 kg of dry matter/cow above a target residual of 1,500 kg of dry matter/ha), but requires confirmation under longer and more complex virtual fencing applications. Within each treatment period, uniform daily pasture utilization (% of pasture consumed above a target residual of 1,500 kg of dry matter/ha) within inclusion zones indicates that cows did not avoid grazing near electric or virtual front-fences. Overall, this study demonstrated a successful simple application of this virtual fencing system to contain a herd of grazing lactating dairy cows within the boundaries of their daily pasture allocation.

Virtual fencing contains and controls grazing cattle using sensory cues rather than physical fences. The technology comprises a neckband-mounted device that delivers an audio cue when the animal nears a virtual boundary that has been set via global positioning system, followed by an electrical stimulus if it walks beyond the boundary. Virtual fencing has successfully been used to intensively graze cattle using a simple virtual front-fence, but a more complex intensive grazing system comprising moving virtual front and back-fences has not been assessed. We studied the effectiveness of virtual fencing technology to contain groups of Angus heifers within grazing cells defined by semi-permanent electric side-fences and virtual front and back-fences, compared to groups of heifers contained in cells defined only by electric fencing. Four groups of 10 Angus heifers were randomly allocated to a “control” (grazed with a conventional electric front and back-fence, n = 2 groups) or “virtual fence” treatment (grazed with a virtual front and back-fence, n = 2 groups). The groups of heifers grazed four adjacent experimental paddocks that were established using TechnoGrazing™ infrastructure. An estimated 9.5 kg pasture DM/heifer.day was offered in each of three 3 day allocations (9 day study period). Data collected include cues delivered by the neckbands, time beyond the virtual boundaries, pasture consumption for each allocation and heifer live weight changes over the study period. The virtual front and back-fences successfully contained one group of heifers in their grazing cell, but the second group of heifers spent an increasing amount of time in the exclusion zone during the second and third allocations and consequently received an increasing number of audio and electrical stimuli. There were no effects of electric or virtual-fence treatment on live weight change or pasture utilization. By grazing heifers in adjacent paddocks our experimental design may have produced a motivation for some heifers to cross the virtual boundary to regain close contact with familiar conspecifics. Despite this, valuable learnings were gained from this study. Most notably, virtual fencing should not be used to manage cattle that have close visual contact to other mobs. We conclude that the successful application of virtual fencing technology needs to accommodate the natural behaviors of cattle.

This study examined the relationships between individual sow aggressive behavior and sow welfare, based on aggression, skin injuries, and stress, in a total of 275 pregnant domestic sows. Over 4 time replicates, sows were randomly mixed into groups of 10 (floor space of 1.8 m²/sow) within 7 d of insemination in both their first and second gestations (200 sows per gestation with 126 sows observed in both gestations). Measurements were taken on aggression (both delivered and received) at feeding, skin injuries, and plasma cortisol concentrations at d 2, 9, and 51 after mixing. Live weight gain, nonreproductive removals, litter size (born alive, total born, and stillborn piglets), and farrowing rate were also recorded. In both the first and the second gestations, sows were classified at d 2 after mixing as “submissive” (delivered little or no aggression at feeding relative to aggression received), “subdominant” (received more aggression at feeding than delivered), and “dominant” (delivered more aggression at feeding than received). In both gestations, sows classified as dominant at d 2 subsequently delivered more (gestation 1, P < 0.01; gestation 2, P < 0.01) and received less (gestation 1, P < 0.01; gestation 2, P < 0.01) aggression and gained the most weight (gestation 1, P < 0.01; gestation 2, P < 0.01). Dominant sows had the least skin injuries throughout gestation 1 (P = 0.04), and although submissive sows sustained the most skin injuries at d 9 and 51 of gestation 2, at d 2 the classifications did not differ in skin injuries (P < 0.01). Subdominant sows had the highest cortisol concentrations at d 2 of gestation 2, but there were no differences between classifications at d 9 and 51 in either gestation (gestation 1, P > 0.05; gestation 2, P = 0.02). There were no significant relationships between aggression classification and reproduction and nonreproductive removals (P > 0.05). In conclusion, sows classified as dominant at feeding at d 2 subsequently received less aggression at feeding, sustained fewer skin injuries, and had higher live weight gain. Submissive and subdominant sows in groups are likely to benefit from the provision of increased resources such as space and access to feed.