Abstract

In-line milk meters are devices which measure the milk yield of dairy cows at every milking; however, they are also susceptible to error. This results in frequent, but inaccurate, measurements. This paper attempts to model a lactation curve and accumulated milk yields from these devices. Wood’s model of a lactation curve was used to predict milk yield. This was modified by adding both systemic and random daily variation. The meters were modelled by taking the modified Wood’s lactation curve for each cow and altering it with various error terms. Along with precision, each sensor also had a rate of drift, and a rate at which its variance grew. It was found that the precision of a sensor can be low, however by aggregating multiple records it is possible to obtain an accurate value for total yield. One strategy that improved the total yield accuracy was to scale the measurements obtained for a particular sensor in a given milking. Due to the uncertainty surrounding the errors involved, and the large assumptions that were made, this model is a good starting point for future work. It is able to identify the relative effect of errors on meter accuracy, and test strategies to improve accuracy of daily and total milk yield.

O, Dowson, G Anderson, and K Hempstalk

Proceedings of the New Zealand Society of Animal Production, Volume 74, Napier, 35-40, 2014