Technical note: Validation of sensor-recorded lying bouts in lactating dairy cows using a 2-sensor approach

Kok, A.; Knegsel, A.T.M. van; Middelaar, C.E. van; Hogeveen, H.; Kemp, B.; Boer, I.J.M. de


Lying behavior is a relevant indicator for the evaluation of cow welfare. Lying can be recorded automatically by data loggers attached to one of the hind legs of a cow. A threshold for the duration of a lying bout (LB) record is required, however, to discard false records caused by horizontal leg movements, such as scratching. Previously determined thresholds for similar sensors ranged from 25 s to 4 min. We aimed to validate LB recorded by the IceQube sensor (with IceManager software; IceRobotics, South Queensferry, UK) and to determine a threshold to distinguish true from false LB records in lactating dairy cows. A novel method of validation, which does not require time-consuming behavioral observations, was used to generate a larger data set with potentially more incidental short LB records. Both hind legs of 28 lactating dairy cows were equipped with an IceQube sensor for a period of 6 d and used as each other's validation. Classification of LB records as true (actual LB) or false (recorded while standing) was based on 3 assumptions. First, all standing records (absence of LB records) were assumed to occur while standing. Second, false LB records due to short leg movements could not occur in both hind legs simultaneously. Third, true LB only occurred if the LB records of the paired sensors coincided. False LB records constituted 4% of the records. Based on a maximum accuracy of 0.99, a minimum duration of LB records of 33 s was determined, implying a sensitivity of 0.99 and a specificity of 0.98. Applying this threshold of 33 s hardly affected estimates of daily lying time, but improved estimates of frequency and mean duration of LB for individual cows. The importance of distinguishing short LB was demonstrated specifically for detection of calving. The 2-sensor approach, using sensor outputs on both hind legs as each other's validation, is a time-efficient method to validate LB records that can be applied to different sensors and husbandry conditions.