Author: Kristine Piccart (ILVO) - Date: April 10, 2019
Ever since their introduction in the 1980’s, automatic milk feeders have a become well-established tool for managing dairy calves in groups. Not only do they reduce labor input, automatic milk feeders also generate a large amount of data on the feeding behavior of individual calves. This makes is easier to monitor their health and wellbeing.
The software of automatic milk feeders can alert users when the feeding behavior of a calf deviates from the norm (for instance, when the milk consumption is lower than expected). Sick calves generally drink less, drink slower, and show less unrewarded visits at the automatic milk feeder (i.e. visits where they do not receive a portion of milk). The impact of disease on the feeding behavior also depends on the feeding regimen: sick calves fed a low milk allowance show no difference in milk intake, as opposed to calves fed “ad libitum”. Furthermore, calves on a restricted milk allowance visit the automatic milk feeder more often and try to suckle more frequently, meaning that a rise in unrewarded visits might actually reflect appetite or hunger. However, the current algorithms built in the software of automatic milk feeders are -generally speaking- not yet good enough to identify sick calves or calves in distress with high accuracy. Most software doesn’t evaluate individual differences or combine multiple parameters – leaving lots of room for improvement.
And who knows, maybe one day we can use the data from automatic milk feeders to predict which calves will turn out to be the most productive cows on the farm (or not)? During a “hackathon” in 2017, a team of data scientists tried to create a mathematical model to predict the future milk yield and composition (i.e. fat, protein) of dairy calves based on data from automatic milk feeders. Unfortunately, the biggest challenge was ultimately the low quality of the data. The milk consumption was oftentimes not registered by the automatic milk feeder, making it impossible to model the data. But first, to unlock the true potential of automatic milk feeders, we need to focus on improving the quality of the data.
Acknowledgements: Thanks to the COST Action initiative DairyCare, the author was able to participate in a short term scientific mission at the Norwegian Veterinary Institute. The title of the work is "Exploring the potential of automatic calf feeder data in the context of keeping cow & calf together".