On-farm welfare assessment of commercial Pekin ducks: a comparison of methods

Abstract

Although a number of welfare assessment methods have been developed for poultry, none have been evaluated for use in commercial duck farms. The primary objective of the study was to evaluate the inter-rater reliability and relative accuracy of 4 duck welfare assessment strategies. Over 2 experiments, 12 flocks of commercial meat ducks (5,850 to 6,300 ducks/flock) aged 30 to 34 D were evaluated. During experiment 1, six flocks were evaluated using 2 welfare assessment methods: transect walks (TW) and catch-and-inspect (CAI). During TW, 2 observers walked predetermined transects along the length of the house and recorded the number of ducks per transect that were featherless, were dirty, were lethargic, had bloody feathers, had infected eyes, and/or had plugged nostrils or were found dead. During CAI, a total of 150 ducks per flock were corralled and individually evaluated. The same welfare indicators were assessed using both methods. During experiment 2, six flocks were initially evaluated using CAI, TW, and a distance evaluation (DE; a total of 50 ducks per flock evaluated from a walking distance) and then reassessed within 24 h during the loadout (LO) process. Data were analyzed in SAS (version 9.4) to determine the observer and method effects on the incidence of welfare indicators. Interobserver reliability was high (P > 0.05) across methods for most welfare indicators. The assessment method affected the measured outcome variables in both experiments (P < 0.05). CAI resulted in higher estimated incidences of most welfare indicators than TW (experiment 1 and 2) and LO (experiment 2). DE yielded intermediate results compared with other methods (experiment 2). Results obtained using TW and LO were most similar, the only difference being the number of dead birds observed using each method (P < 0.0001). The average time required for CAI, TW, DE, and LO was 2.40 ± 0.004, 1.12 ± 0.02, 1.54 ± 0.001, 3.56 ± 0.006 h, respectively. Bootstrapping analyses showed that the observed welfare indicator prevalence estimates were affected by the number of transects (TW) and number of birds (CAI) sampled.

Keywords: Pekin duck; individual sampling; loadout; transect walks; welfare assessment.

Figure 1

Each house was divided longitudinally into 7 transects from wall to wall. Two observers walked the length of each transect in a semirandom (evaluation of adjacent transects was avoided) order that was determined independently for each flock and recorded the number of birds showing any of the predefined welfare problems. During the walk, the observers score the indicators in a zone covered by a semicircular area, 1 m in front of the observer. The location of drinkers and feeders marked the edges of the transects, whereas the red line delineates the path traveled by the observers. A sample transect sampling order is provided in red numbers.

Figure 2

Mean values and SEM of the data set for bloody feather (A) and footpad score 1 (B) were calculated by taking random samples of individual evaluation of 25, 50, 100, and 150 birds. Simulations were run 10,000 times using bootstrapping technique in R.

Figure 3

Mean values and SEM for gait score 2 and (A) blood on feather (B) expressed as percentage for 2, 3, 4, 5, and 7 transects used in 10,000 simulation using bootstrapping technique in R.