The Udder Zone: using zone-based behavioural scoring during nursing bouts to predict individual fallback risk in neonatal piglets without continuous tracking

Authors:

Abstract
In intensive pig production, piglets face high competition for functional teats, which can lead to malnutrition and pre-weaning mortality. Behavioural observations of suckling offer a non-invasive method to identify piglets falling back in growth (i.e., ’fallbacks’) early on. However, tracking of these piglets is difficult due to frequent occlusions by farm equipment and litter mates. To avoid the constraints imposed by occlusions, we explored whether zone-based behavioural scoring can characterise individual piglet suckling behaviour and predict fallback risk, without continuous tracking. We monitored five sows and their litters in farrowing pens equipped with cameras throughout early lactation. We weighed and individually marked piglets during processing, and again on days 2, 6, and 14 post-partum, from which we calculated individual average daily weight gain (ADG) for each interval. Piglets that died or fell within the lowest 15% of ADG within each litter × weighing interval were classified as fallbacks. For each of the five litters, 24 nursing bouts were analysed: the first eight following the first three weighings – at processing, and on days 2 and 6. Nursing bouts were analysed using an ethogram including: ’in Udder Zone’, ’walking’, ’lying’, or ’out of sight (not in Udder Zone)’. The Udder Zone was defined as a rectangular area surrounding the sow’s udder from her anterior shoulder to her posterior shoulder and one piglet body length wide. From the observations, we calculated metrics summarising bout participation, initiation, and teat fidelity. Principal component analysis reduced behavioural metrics to one dominant engagement axis (PC1, explaining 35.6% of variance), characterised by high bout participation and short latencies to join. This engagement axis was negatively associated with fallback probability, with higher PC1 scores (reflecting increased and stable nursing engagement) at the interval start corresponding to reduced fallback probability at the end of the interval (OR = 0.48, 95% CI: 0.31–0.73, p < 0.001). Fallback probability was higher over the D6-D14 interval (estimated marginal probability: 26.4%, 95% CI: 8.1–59.3%) than over processing-D2 (1.4%, 95% CI: 0.2–7.8%, p = 0.009) and D2-D6 (2.9%, 95% CI: 0.6–13.3%, p = 0.029), consistent with a critical window around teat-order stabilisation after which less engaged piglets are unlikely to recover. Taken together, the results provide observational evidence that individual zone-based behavioural scoring (without requiring continuous tracking) during nursing bouts captures meaningful information about piglet growth performance and survival. Such indicators could form the basis for future automated monitoring systems.