Technical note: The use of a physical activity monitor to estimate the eating time of cows in pasture

Abstract

Accurate eating time can be used as an index of forage dry matter intake in grazing cows. To develop a method for easily estimating the eating time of dairy cows in a pasture, 8 lactating Holstein cows were fitted with collars equipped with commercial uniaxial accelerometers; namely, the Kenz Lifecorder EX (LCEX; Suzuken Co. Ltd., Nagoya, Japan), and were allowed to graze in a pasture for 4, 8, or 20 h daily for 7 d. The LCEX device recorded the intensity of the physical activity categorized into 1 of 11 activity levels ranging among 0 (no movement), 0.5 (subtle) and from 1 to 9 (1, light; 9, vigorous intensity) every 4s during the experimental period. The activities of the animals were also video-recorded for 11h and were manually classified into 7 categories (eating, searching, ruminating, standing resting, lying resting, drinking, and walking) at 4-s intervals. According to the count distribution of the activity levels for the categorized activities, 94.4% of the counts involving eating activity ranged from activity level 1 to 7. On the other hand, most of the counts were activity level 0 or 0.5 when ruminating and resting activities were observed. No records of activity level 8 or 9 were found in any activities. When activity level 1 was used as a threshold for discriminating eating from the other activities, the lowest misclassification rate of 5.5% was observed. With a threshold of activity level 1, the eating times in pasture for cows grazing for 4, 8, and 20 h/d were 142.8, 290.6, and 438.4 min/d, respectively, and the proportions of the time spent in pasture that were made up of eating time were 0.66, 0.67, and 0.38, respectively [the proportion during daytime (8h of the 20-h grazing treatment) was 0.63 and that at nighttime (12h of the 20-h grazing treatment) was 0.23]. The use of the LCEX device allows for easy measurement of eating time and facilitates the determination of the pattern of eating activity in pasture for grazing cows.

Behavioral classification of data from collars containing motion sensors in grazing cattle

“…Therefore, the use of both head position and activity level, and the higher frequency of data collection in our study seem to improve the precision (by 2-13% units) and sensitivity (by 10-16% units) of classification of foraging behavior compared to previous studies which could make the methodology and electronic technology developed in the present study more appropriate to identify and interpret foraging hotspots. Ueda et al (2011) achieved 94.5% for both precision and specificity for grazing vs. 'all other behaviors' using discriminant analysis of uni-axial accelerometer data in dairy cows grazing small paddocks with temperate pastures. Martiskainen et al (2009) classified 6 activities using support vector machines and data from collars containing 3-axis accelerometers in dairy cows housed in a barn.…”

Daily grazing time of dairy cows is recorded accurately using the Lifecorder Plus device

“…They might not be sufficiently accurate for scientific use, however, because ruminant nutrition research requires a greater level of precision at short-or medium-time scale. Recently, a portable device called Lifecorder (Ex version, LCP, Suzuken Co. Ltd., Nagoya, Japan), originally designed for human health improvement, has been suggested by Ueda et al (2011) as a potentially useful device for recording the grazing behaviour of dairy cows at pasture. Lifecorder is based on an uniaxial accelerometer that records physical activity level (range 1-9) for each 4-s period.…”

“…Following the work of Ueda et al (2011) and Yoshitoshi et al (2013), the objective of this experiment is to determine the accuracy of the Lifecorder Plus device in recording the grazing activity periods of dairy cows at pasture more extensively. The novel aspects of this study are based on (1) the use of the Plus version of the Lifecorder that potentially does not require preliminary data processing to identify grazing activity periods, contrary to the Ex version, (2) the duration of the validation period, of more than 200 h compared to the 11-h validation period by Ueda et al (2011) and to the 15-h validation period by Yoshitoshi et al (2013), and (3) the comparison of activity levels recorded when the device is placed either on the neck or on the leg.…”

Dairy cattle behavior classifications based on decision tree learning using 3‐axis neck‐mounted accelerometers

“…Therefore, an automated behavior‐monitoring system is needed. In response to this need various research studies have been conducted and published, such as: a one‐axis accelerometer mounted to the collar of cattle (Ueda, Akiyama, Asakuma, & Watanabe, ), the detection of eating and estimation of feed intake by pendulum (Uemura, Wanaka, & Ueno, ), a bitemeter incorporated microphone and a mercury switch (Delagarde, Caudal, & Peyraud, ), a pressure sensor attached to the halter (Braun, Trösch, Nydegger, & Hässig, ), the detection of eating and rumination by one‐axis accelerometer with a voice recorder attached to a horn (Tani, Yokota, Yayota, & Ohtani, ), the detection of rumination time by a logger attached to the collar (Schirmann, von Keyserlingk, Weary, Veira, & Heuwieser, ), and the monitoring of lying behavior using a pedometor (Mattachini, Antler, Riva, Arbel, & Provolo, ). However, these are limited to the detection of simple behaviors, which is unsuitable for understanding complex behaviors.…”