The behaviour and movement of lame dairy cows at pasture have been studied little, yet they could be relevant to improve the automatic detection of lameness in cows in pasture-based systems. Our aim in this study is to identify behavioural and movement variables of dairy cows at pasture that could discriminate lameness scores. Individual cow behaviours were predicted from accelerometer data and movements measured using GPS data. Sixty-eight dairy cows from three pasture-based commercial farms were equipped with a 3-D accelerometer and a GPS sensor fixed on a neck collar for 1-5 weeks, depending on the farm, in spring and summer 2018. A lameness score was assigned to each cow by a trained observer twice a week. Behaviours were predicted every 10 s based on accelerometer data, and then combined with the GPS position. Segmentation on behavioural time series was used to delineate each behavioural bout within each outdoor period. Thirty-seven behavioural and movement variables were then calculated from the behavioural bouts for each cow. A partial least square discriminant analysis was performed to identify the variables that best discriminate lameness scores. Time spent grazing, grazing bout duration, duration before lying down in the pasture, time spent resting, number of resting bouts, distance travelled during grazing, and dispersion were the most discriminant variables in the PLS-DA (VIP > 1). Severely lame cows spent 4.5 times less time grazing and almost twice as much time resting as their sound congeners, especially in the lying position. Exploratory behaviour was also reduced for both moderately and severely lame cows, resulting in 1.2 and 1.7 times less distance travelled respectively, especially during grazing. These variables could be used as additional variables to improve the performance of existing lameness detection devices in pasture-based systems.
Keywords: Accelerometer; Behaviour and movement; Dairy cow lameness; GPS; PLS-DA; Pasture.
Copyright © 2021 Elsevier B.V. All rights reserved.