| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Animal Welfare Program, Faculty of Agricultural Sciences, The University of British Columbia, Vancouver V6T 1Z4, Canada
Agriculture and Agri-Food Canada, Research Centre, Lethbridge, Alberta T1J 4B1, Canada
Corresponding author: T. DeVries; e-mail: trevorjd{at}interchange.ubc.ca.
 |
ABSTRACT |
|---|
 TOP ABSTRACT ACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: feeding behavior • dairy cow • validation
The time spent eating, and the pattern of meals, can obviously have important effects on total daily intake of dairy cattle (Grant and Albright, 2000). For that reason, a great deal of recent research in dairy nutrition and management has focused not only on changes in intake, but also on changes in feeding behavior. Over the past few decades, management of North American dairy farms has moved from tie stalls to loose housing systems such as free stalls. This move to loose housing has made the collection of feeding behavior data more difficult. Traditionally, feeding behavior of loose-housed cows was monitored through direct human observation, but this method is labor intensive, making observations on multiple animals and days difficult. The use of time-lapse video recordings (Friend et al., 1977; Vasilatos and Wangsness, 1980) increases the ability to monitor many animals for multiple days; however, transcription of these videos is also labor intensive.
With the introduction of various individual intake monitoring systems, researchers (e.g., Tolkamp et al., 2000) have collected continuous feeding behavior data for loose-housed cows. However, cows using these systems are required to access feed via an electronic gate, potentially changing feeding behavior compared to that in commercial loose-housed systems. Access to feed via a specific feeding station may also reduce the number of competitive interactions that occur during feeding compared with feeding from an open feed alley like that used on commercial loose-housed farms.
Radio frequency has been employed for many years as a means of electronically identifying cows (Eradus and Jansen, 1999). Recently, a radio frequency electronic monitoring system (GrowSafe Systems Ltd., Airdrie, AB, Canada) has been designed that allows for the passive monitoring of individual cow presence at the feed alley. The system was originally described by Sowell et al. (1998) and validated by Schwartzkopf-Genswein et al. (1999) for the monitoring of feeding behavior of feedlot cattle, but this technology has not been validated for measuring feeding behavior of loose-housed dairy cattle fed via a feed alley. The objective of this study was to validate the measures generated by the GrowSafe system, by comparing these estimates with measures from time-lapse video recordings.
Two groups of six lactating Holstein cows were monitored continuously for 24 h using the GrowSafe feed alley monitoring system and time-lapse video. The cows were housed in adjacent pens in a free-stall barn located at The University of British Columbia Dairy Education and Research Centre (Agassiz, BC, Canada) and were managed according to the guidelines set by the Canadian Council on Animal Care (1993). Each animal had access to a sand-bedded free stall. Cows were fed a TMR (30% corn silage, 8% grass silage, 4% alfalfa hay, 5% third-cut grass hay, 16% steam-rolled corn, and 37% concentrate mash on a DM basis) from a feed alley (0.6 m of space/cow) with access via a neck rail. Animals were fed daily at approximately 0600 and 1600 h and were milked at approximately 0600 and 1700 h daily.
The GrowSafe system was modified from that previously described (Schwartzkopf-Genswein et al., 1999), so that data transfer from the reader panel to the computer occurred via radio frequency. The antenna mats (each 7.2 m long and 0.75 m wide) used to detect transponders were laid on the floor of the feed alley adjacent to the tombstone with feed delivered on top of the mats. All animals were fitted with a passive transponder, which was encased in a plastic ear tag (All Flex, Inc., Dallas, TX) and attached to the bottom of the neck collar. The system was designed to detect the transponder when within 50 cm of the antenna mat, such as when a cow placed her head under the neck rail and over the feed. This system worked by scanning set locations along the mat each at 6-s intervals. Whenever a transponder was within range, the reader panel recorded presence (a ‘hit’) of the transponder including time and location on the mat. These data (transponder number and time stamp) were downloaded continuously via radio frequency to a computer housed approximately 100 m from the panel. The computer was equipped with GrowSafe feed alley monitoring software version 6.38.
A video camera (Panasonic WV-BP330; Osaka, Japan) was positioned approximately 6 m above the feed alley of each experimental pen. Output from the cameras was recorded with a time-lapse video recorder (Panasonic AG-6540) in 48-h mode and a digital video multiplexer (Panasonic WJ-FS216). The clocks on the video recorder and the computer collecting the GrowSafe data were synchronized at the outset of the experiment. Red lights (100 W) were hung 6 m above the feed alley to facilitate video recording at night. Cows were individually identified with symbols on both sides of their body using hair dye. Cows were scored as present at the feed alley when the neck collar passed over the tombstone and above the feeding surface.
Raw data were summarized for each cow and for each minute of the day (n = 1440), recording cow presence (1) or absence (0) at the feed alley as determined by both the GrowSafe system and video.
Meal-based estimates of feeding behavior rely on the use of a meal criterion, which is the minimum time interval between visits to consider the next feed alley visit as being part of a new meal. A meal criterion of 27.74 min was used following DeVries et al. (accepted). This criterion was used to calculate meal frequency (meals d-1) by counting the number of intervals between visits to the feed alley that exceeded the criterion and adding one. Meal duration (min meal-1) was calculated as the time from the first visit until, but not including, an interval between visits that exceeded the criterion.
We regressed (SAS, 1985) measures of behavior generated by GrowSafe (dependent variables) onto those from video (independent variables), testing for slope and intercept effects. Comparisons between the video and GrowSafe animal presence data within cows were performed by chi-square 2 x 2 contingency tables (presence and absence at each minute, as determined by GrowSafe and video). These data were also used to calculate predictability (likelihood that a cow detected by GrowSafe is actually present at the feed alley), sensitivity (likelihood that a cow present at the feed alley is detected present by GrowSafe), and specificity (likelihood that a cow that is absent from the feed alley is detected as absent by GrowSafe) estimates according to Martin et al. (1987).
The GrowSafe monitoring system provided estimates of feeding behavior that were very similar to that observed from video. Meal durations ranged from 31.72 to 57.49 min meal-1 for the GrowSafe data and 33.42 to 58.37 min meal-1 for the video data. Meal frequency ranged from 5 to 10 meals d-1 for both the GrowSafe and video data. The line equations and coefficients of determination were calculated for meal duration (y = 0.96x + 0.63, r2 = 0.98) and for meal frequency (y = x, r2 = 1). In both cases the coefficients were significant (P < 0.001), but the intercept did not differ from zero and the slope did not differ from one (P > 0.3).
In the previous validation work with this system using beef feedlot cattle, researchers found that the GrowSafe data for meal frequency and meal duration were overestimated compared with that found from video analysis (Schwartzkopf-Genswein et al., 1999). However, these researchers used a short meal criterion (85 s) that had not been established objectively. The use of an objectively calculated meal criterion (Tolkamp et al., 2000; DeVries et al., 2003) likely results in a more biologically relevant estimate of meal measures. In the current study we used an objectively defined meal criterion and found a very good agreement between the GrowSafe meal measures and those from video.
The monitoring system records presence of individual cows at the feed alley. Recent work with the GrowSafe system in a free-stall dairy barn has shown that the most repeatable measures across time were based on the number of hits (measure of presence) at the feed alley: a measure that does not rely on a meal criterion (DeVries et al., 2003). However, prior to the current study, there has been no published work to validate the GrowSafe system’s ability to monitor animal presence at the feed alley. Figure 1
illustrates the agreement of estimates of animal presence as collected by both GrowSafe and video. For all 12 cows there was a high correlation (Table 1) between the two methods in detecting presence at the feed alley. High values for predictability (96.5%), sensitivity (87.4%), and specificity (99.2%) all indicate that the GrowSafe system is a reasonable method for detecting cow presence.
|
|
Additionally, for 3.5% of observations when the GrowSafe systems indicated that a cow was present at the feed alley, the video showed that the cow was not present. These extraneous observations were likely due to interference caused by the physical structures of the facility. For example, there were several physical structures (e.g., gates, fencing, lying stall partitions, and components of the feed alley neck rail) that may have acted as unintended antennae for signals (Schwartzkopf-Genswein et al., 1999). As illustrated in Figure 1
, even for the cow with the worst correlation (cow 9) between the two data collection methods, all feeding observations by GrowSafe occurred during feeding bouts confirmed by video. This result suggests that it was likely the metal near the feed alley (i.e., neck rail) that acted as a false antenna when cows were in close proximity to the feed alley.
Although we have validated the GrowSafe records against the time-lapse video observations, some errors may also occur with video observations. For example, when using time-lapse video, errors may occur in recording the exact times when a cow’s neck collar came within the read range of the mat. Additionally, due to the 6 s scanning method of the GrowSafe system, some animals may not be detected immediately when they start to feed, resulting in a different entry time compared with the video data.
In conclusion, the GrowSafe system can be used to monitor the feeding behavior of loose-housed dairy cattle. The GrowSafe system, used with an appropriate meal criterion, provides an accurate estimation of meal-based measures of feeding behavior. Furthermore, this system provides a reasonable estimate of when animals are present at the feed alley. Recording errors can be reduced by carefully monitoring factors such as transponder position and the number of metallic objects within close proximity of the feeding area.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication June 18, 2003. Accepted for publication August 8, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTACKNOWLEDGEMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. Chapinal, D. M. Veira, D. M. Weary, and M. A. G. von Keyserlingk Technical Note: Validation of a System for Monitoring Individual Feeding and Drinking Behavior and Intake in Group-Housed Cattle J Dairy Sci, December 1, 2007; 90(12): 5732 - 5736. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Urton, M. A. G. von Keyserlingk, and D. M. Weary Feeding Behavior Identifies Dairy Cows at Risk for Metritis J Dairy Sci, August 1, 2005; 88(8): 2843 - 2849. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Huzzey, M. A. G. von Keyserlingk, and D. M. Weary Changes in Feeding, Drinking, and Standing Behavior of Dairy Cows During the Transition Period J Dairy Sci, July 1, 2005; 88(7): 2454 - 2461. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. J. DeVries and M. A. G. von Keyserlingk Time of Feed Delivery Affects the Feeding and Lying Patterns of Dairy Cows J Dairy Sci, February 1, 2005; 88(2): 625 - 631. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Bach, C. Iglesias, and I. Busto Technical Note: A Computerized System for Monitoring Feeding Behavior and Individual Feed Intake of Dairy Cattle J Dairy Sci, December 1, 2004; 87(12): 4207 - 4209. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. J. DeVries, M. A. G. von Keyserlingk, and D. M. Weary Effect of Feeding Space on the Inter-Cow Distance, Aggression, and Feeding Behavior of Free-Stall Housed Lactating Dairy Cows J Dairy Sci, May 1, 2004; 87(5): 1432 - 1438. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
