# Assessing the Reliability of Automatic Milking Systems Data to Support Genetic Improvement in Dairy Cattle

## TL;DR

This study evaluates how reliable data from automatic milking systems is for improving dairy cattle breeding through genetic analysis.

## Contribution

The study provides novel insights into the genetic reliability of traits measured by automatic milking systems compared to official milk recording data.

## Key findings

- Automatic milking systems data showed high correlation with official data for milk yield but lower for fat and protein percentage.
- Heritability estimates from automatic milking systems were higher for production traits compared to official data.
- Systematic deviations between systems suggest the need for calibration, especially for quality traits.

## Abstract

Automatic milking systems are rapidly spreading among dairy farms and represent a valuable source of phenotypes thanks to the continuous monitoring of numerous traits. This real-time monitoring could enable a more accurate characterization of phenotypes and, consequently, more precise breeding. In this study, we evaluate the potential of using these data for selection by estimating genetic parameters, such as heritability, both for traits related to production and resistance to mastitis. Findings suggest that milk yield and somatic cell traits measured by automatic milking systems can be reliable and other traits could gain the advantage of appropriate calibration using external data.

This study investigates the reliability and potential genetic utility of data recorded by automatic milking systems by comparing them with official milk recording data. Analyses focused on phenotypic distributions, correlations, systematic differences, and heritability estimates for milk production and quality traits including milk yield, fat and protein percentage, somatic cell count, and electrical conductivity. Automatic milking system data and official milk recording data shared similar distributions. Correlations between the two systems were high for milk yield (r = 0.93), but moderate for fat (r = 0.52) and protein percentage (r = 0.48), and somatic cell count (r = 0.62), suggesting that while the former provides consistent data for quantity traits, quality-related ones may be less reliable. Systematic deviations between automatic milking systems and official milk recordings emerged across different lactation stages. Heritability estimates based on automatic milking system data were generally higher than the official control for production traits, supporting their use in genetic evaluations. Electrical conductivity displayed a similar heritability to somatic cell count, but its measure is insufficiently detailed and its use as an indirect indicator of udder health is not recommended. Automatic milking system data demonstrates potential for integration into genetic selection programs, although further refinement of sensor accuracy is recommended.