The assessment of rising and lying down behaviours of dairy cows in cubicle housing systems with regard to animal welfare

Abstract

In cubicle housing systems for dairy cows, the lying area is divided into individual lying places called lying cubicles. In these cubicles, contamination of the bedding is limited to reduce resource and labour costs and to promote animal hygiene and health. However, lying cubicle design restricts cow behaviour and limits their freedom of movement. This can be particularly problematic as cows rise to the standing position and lie down according to innate, species specific movement patterns with limited ability to adapt these movements to their environment. Inadequate movement space can result in atypical rising and lying down behaviours and possibly a reduced lying frequency. However, the direct effect of lying cubicle design on rising and lying down movements is a relatively understudied aspect of dairy cow welfare. Therefore, the aim of this thesis was to gain a deeper insight into the effects of lying cubicle design on the rising and lying down behaviours of dairy cows with regard to animal welfare. Firstly, an automated method to detect specific atypical rising and lying down behaviours in dairy cows was proposed. Cows were fitted with four tri-axial accelerometers attached to both front legs, the left hind leg, and the head. Acceleration time series were matched with presence/absence labels for atypical rising and lying down behaviours from video observations, after which various machine learning algorithms were used for model development. Next, the association between lying cubicle design and the quality of rising and lying down movements was investigated on commercial Swiss dairy farms with small lunge spaces. Farms had either ‘permissive’ lying cubicles with open frame partitions that facilitate lateral space sharing and flexible neck straps, or ‘restrictive’ cubicles with partitions with more bar work and rigid neck rails. The prevalence of atypical rising and lying down behaviours was determined from video recordings. Lastly, the effect of neck strap positioning on rising and lying down behaviours and cubicle hygiene was investigated. Three neck strap heights and two distances from the curb were examined consecutively in an experimental barn with two mirrored pens. Rising and lying down behaviours and elimination behaviour around rising events were analysed from video recordings and cubicle cleanliness was assessed for photographs made at afternoon milking. Atypical rising and lying down behaviours were detected with balanced accuracies of 0.56 to 0.74. This is not yet satisfactory for use in the evaluation of dairy cow housing systems, but subsequent analysis indicated that performance may improve with more training data. Ethograms designed for human observers are likely suboptimal for machine learning, and adjustments may be needed with machine learnability in mind. In the permissive cubicle type, staggered head lunges and displays of hesitation before lying down were less prevalent, lying frequency was higher and lying time was longer. Soiling of the bedding was limited with flexible neck straps positioned restrictively in terms of recommendations for neck rail placement, while rising and lying down movements or general lying behaviour were not considerably affected. The results suggest that open frame partitions and flexible neck straps can improve the conditions for cows to cope with an atypical environment and promote the expression of natural rising and lying down behaviours, while ensuring the cleanliness and health of the cow.