Spatial behaviour of dairy cows is affected by lameness

Lameness is one of the major welfare problems on modern dairy farms, and additionally, it is difficult to control. Lameness is associated with changes in cow behaviour, and efforts have been made to automatically detect these behavioural changes. However, systems relying on a single behavioural variable are likely to fail. Indoor positioning could provide means to measure multiple behavioural variables with a single system. Our aim was to investigate how lameness affects the spatial behaviour of cows, measured with an indoor positioning system. In total, 71 lactating dairy cows were followed during a 7-month study period, with 48 cows in the study simultaneously. Cows were locomotion scored fortnightly with a 10-tier scale, and their daily time spent in the different functional areas of the barn, walking distance, and home range were calculated from the positioning data. Each locomotion score was merged with the 5-day average of the behaviour variables leading up to the scoring day, resulting in 376 observations in the final data. Linear mixed models were fitted with backwards stepwise elimination to test the associations between positioning-based daily behavioral variables and predictor variables comprising locomotion score, parity, lactation stage, breed and the proportion of missing positioning data. Increasing locomotion score was associated with increased time spent in the lying stalls (P = 0.0037) and decreased time spent in the alley (P < 0.0001). Positioning-based feeding time was confounded by parity (P = 0.011) as the model used to estimate the feeding time from the position data was less sensitive in classifying primiparous cows correctly as feeding or not feeding. Severe lameness was also associated with a shorter daily walking distance (P = 0.0447) and smaller core home range (P = 0.005). Proportion of missing positioning data affected only daily walking distance (P < 0.0001) and full home range (P = 0.0059), and distance-based variables seemed more sensitive to data quality compared to spatiotemporal variables. Our results show that indoor positioning of dairy cows has a potential to contribute to development of automatic lameness detection. However, reliability of positioning systems should be improved, and the amount of missing data should be minimised to improve the calculation of distance-based variables