Last week I attended the 2017 American Dairy Science Association meeting in Pittsburgh, PA. I presented a poster on our project “Sampling strategies for dairy cow welfare assessments“, which is nearly ready to submit for publication.
During her M.Sc., Julia Trieb visited from Austria and collected about two dozen animal-based measures on 10 California drylot and freestall dairy operations. She restrained the cows in headlocks at the feed bunk to assess measures such as discharge and skin injuries, and released the cows one-by-one to watch them walk in order to score lameness.
Scoring all cows for so many measures can be time consuming, so we wanted to evaluate the accuracy of assessing a subsample of the pen.
We selected proportions ranging from 1/10 to 3/4 of the high-producing pen, and found that in our sample, we had to score at least 2/3 of the pen for all measures to be considered accurate.
We also compared different methods for selecting cows, and we found that choosing cows based on their position while locked at the feed bunk (i.e., pick 2, skip 1, pick 2, skip 1…) performed about the same as selecting cows randomly. Selecting cows based on their lockup position may be a more straightforward approach on farms that have headlocks.