Technical note: Comparison of instantaneous sampling and continuous observation of dairy cattle behavior in freestall housing

Abstract

Recording behavior at fixed intervals (instantaneous sampling) can reduce labor relative to observing continuously. However, instantaneous sampling may inaccurately estimate potentially important responses, such as how frequently cows perform a behavior (i.e., the number of bouts). Our objective was to validate the use of instantaneous sampling for capturing how long and how frequently cows in freestall housing lie down or visit the feed bunk and water trough. We predicted that more frequent sampling would be needed to accurately reflect the behaviors that cows spent less time performing. In addition, we predicted that instantaneous sampling would underestimate how often cows engaged in behaviors that they frequently performed in short bouts or with short intervals between bouts, as some of these events may occur between sample intervals. Continuous video observations of 18 lactating Holstein-Friesian dairy cows were conducted for 48-h periods. Instantaneous samples (1 and 30 s, and 1, 3, 5, 10, 15, and 30 min) were generated from continuous data, with the samples recorded at 1-s intervals representing true values. Estimates from each sample interval ≥30 s were compared pairwise to true values with regression analysis. Sample intervals were considered accurate if they met 3 criteria: coefficient of determination ≥0.9 (i.e., strongly related to true values), slope=1, and intercept=0 (i.e., did not over- or underestimate true values). The amount of time cows spent lying (12.1±1.8h/24h, mean ± standard deviation) or visiting the water trough (1.1±0.8h/24h) and feed bunk (5.6±0.8h/24h) were accurately captured using sample intervals ≤30, 10, and 5 min, respectively. In addition, sample intervals ≤3 min accurately estimated the number of lying bouts (10.3±2.4 per 24h), likely because cows were recumbent for long periods (74.0±17.4 min, on average, with <6% of bouts lasting <5 min) and rarely resumed lying soon after standing up (0.4% of intervals between lying bouts were <30 s). However, shorter sample intervals may be needed in situations where cows more frequently transition between lying and standing. In contrast to lying in this study, cows visited the water trough and feed bunk for shorter periods (3.5±1.7 and 25.6±5.8 min, respectively) and frequently returned to these resources soon after leaving (17 and 7% of intervals between visits were <30 s long). As some of these events likely occurred between sample intervals, all sample intervals ≥30 s underestimated the number of times cows visited the water trough and feed bunk (18.5±6.2 and 14.1±4.4 per 24h, respectively). Therefore, continuous observation is needed to determine how often cows visit these resources.

Keywords: behavior; instantaneous sampling; methodology; validation.