Automated recording of cow brush visits in a commercial dairy farm setting

Farmers often install automatic cow brushes to promote grooming behaviour, potentially reducing stress. Health problems in cattle are typically accompanied by a suit of sickness behaviours and a reduction of low resilience behaviours such as grooming. Thus, decreased automatic brush use could be a potential indicator of disease. Our study aimed to develop and validate an algorithm for automatic monitoring of cow brush usage in a commercial dairy farm setting. The research took place on a commercial dairy farm in the Netherlands housing 130 Holstein Friesian dairy cows fitted with a Nedap SmartTag Neck that included cow location. Visual observations of cow brush usage were performed for 38 hours, distributed across 12 days by two observers, yielding 533 visits to the brush. Cows brushing (87.4% of visits) had a median brushing time of 1:22 minutes (range 00:10-20:03). An algorithm was developed and then validated to determine the time spent at the brush based on location data. Results show good precision (89.1%), recall (87.4%), and F1 score (88.3%) for the algorithm. Time spent at the brush for observations and algorithm was strongly correlated for the true-positives (Spearman’s rank-order correlation: r=0.919; p<0.001; n=466), as were time observed at the brush and brushing time (Spearman’s rank-order correlation: r=0.853; p<0.001; n=533). Our algorithm had a moderate predictive value for brushing time (R2 = 0.409; p<0.001) indicating a need for further optimization. This study is the first step in validating an algorithm for automated recording of brushing time, enabling future studies relating brushing time to health and welfare.