Heritabilities, genetic and phenotypic correlations between body conformation and milking traits of Holstein-Friesian cows milked by milking robots

1. Introduction

Functional traits, including body conformation traits, constitute an important component of selection indices used in dairy cattle breeding programs in countries with advanced genetic evaluation systems. In Poland, for example, conformation traits account for approximately 25% of the total selection index value (PFHBiPM, 2025). Their inclusion is justified by their documented influence on milk production, health, fertility, and longevity in dairy cows (Stefani et al., 2018; Zink et al., 2011; Gibson & Dechow, 2018; Török et al., 2021).

Numerous studies have demonstrated that body conformation traits, particularly those related to udder morphology, are genetically and phenotypically associated with milk production and functional performance. Favorable udder characteristics, such as shallow udders and strong attachment, are generally linked with higher milk yield (MY), improved udder health, and lower somatic cell score (Dadpasand et al., 2012; Morek-Kopeć & Zarnecki, 2012). Reported genetic correlations (r _G) between conformation and MY, however, vary in magnitude and direction, indicating a complex relationship between these traits (Sartori et al., 2018; Kelm & Freeman, 2000; Short & Lawlor, 1992). In addition, specific conformation traits, such as pelvic structure, may influence other functional traits, including calving ease (Nogalski et al., 2024).

Traditionally, cattle breeding programs have prioritized milk production traits; however, numerous studies have identified genetic correlations between conformation features and MY (De Haas et al., 2007; Ptak et al., 2011; Loker et al., 2012; Bohlouli et al., 2015). Greater stature has frequently been associated with increased MY. Body conformation, particularly udder morphology, strongly affects udder health and somatic cell score (SCS). Cows with shallower, well-attached udders and balanced body conformation tend to have lower SCS and reduced susceptibility to mastitis, making conformation traits useful indirect indicators of udder health (Dadpasand et al., 2012).

Recent studies have confirmed a strong connection between conformation and both milk production and reproduction (Battagin et al., 2013; Bobbo et al., 2019; Carthy et al., 2016; Eaglen et al., 2013). These relationships underline the importance of conformation traits as functional indicators that influence overall reproductive efficiency in dairy cattle.

Despite this well-established role of conformation traits in conventional production systems, relatively little is known about their relationship with traits specific to automatic milking systems (AMSs). Since their introduction in the early 1990s, AMSs have been increasingly adopted, particularly in Europe, due to their potential to reduce labor requirements and improve herd management efficiency (Jacobs & Siegford, 2012; Nixon et al., 2009). In addition to these operational benefits, AMSs provide detailed, high-frequency data on milking processes, including traits such as number of attachments (AN), attachment time (AT), and milking efficiency (ME), which may serve as novel indicators of cow adaptability and performance in automated environments (Aerts et al., 2022). While AMS have been associated with increased milking frequency and, in many cases, higher MY, the genetic relationships between AMS-derived traits and traditional conformation characteristics remain insufficiently explored. Understanding these relationships is essential for optimizing breeding strategies in herds utilizing automated technologies. Therefore, the aim of this study was to estimate heritabilities and genetic and phenotypic correlations (r _P) between MY, ME, AN, AT, and selected udder- and non-udder-related conformation traits in Holstein-Friesian cows milked using automated milking systems.

2. Material and methods

The research material consisted of data from 796 primiparous Polish Holstein-Friesian cows kept in 7 herds equipped with Lely Astronaut A4 milking robots (Lely Holding B.V., Maassluis, the Netherlands). The pedigree database was compiled up to three generations and included a total of 3,690 animals. The cows included in the analysis (n = 796), evaluated for conformation and AMS-related milking traits, were daughters of 320 sires, resulting in an average of 2.5 daughters per sire. The studied cows first calved in 2012–2014 to ensure data consistency and comparability across herds and recording conditions. The cows were housed in free-stall barns and fed a partial mixed ration feed. The cows received a variable dose of the concentrate, either from the milking robot or the feeding station, depending on their MY.

Data on the assessment of conformation came from the IT resources of the Polish Federation of Cattle Breeders and Dairy Producers. The linear conformation traits, collected in accordance with the International Committee for Animal Recording (ICAR, 2018) standards, included:

• udder depth (UD, pts),

• teat length (TL, pts),

• rear udder width (RUW, pts),

• rear udder height (RUH, pts),

• rear teat placement (RTP, pts),

• front teat placement (FTP, pts),

• fore udder attachment (FA, pts),

• central ligament (CL, pts),

• angularity (AY, pts),

• foot angle (FTA, pts),

• rear legs rear view (RLRV, pts),

• height at sacrum (HS, cm),

• body depth (BD, pts),

• chest width (CW, pts),

• rump angle (RA, pts),

• rump width (RW, pts),

• rear legs set (RLS, pts).

Primiparous cows were monitored for the following features characterizing the course of automatic milking:

• Daily MY (kg/day) – daily MY of the cow summed during 1 day in milk,

• ME (kg/min) – MY per day divided by box time (total time spent by a cow per day in the robot milking box),

• AN (n/milking) – the average number of attempted connections of the milking device to the teat,

• AT (s/milking) – the average time of attempted connections of the milking device to the teat.

Data on milk performance characteristics of primiparous cows milked in AMS were obtained from the T4C management and data registration system (Lely East). The amount of recorded information from the AMS database for each cow was limited to the 30 days period preceding and the 30 days period following the day of conformation assessment. These data included daily MY and milking parameters. For each animal, the values recorded within this 60 days interval were averaged, and the arithmetic mean values were used for subsequent exploratory and genetic analyses. Records showing milk-performance traits beyond three standard deviations were discarded prior to analysis. The use of 60 days averaged AMS data was necessary due to the structure of the dataset, where conformation traits were recorded as a single observation per cow, while milking parameters were available as repeated daily measurements. Without aggregation, this would result in pseudo-replication. However, this approach may reduce phenotypic variance and lead to partial loss of temporal information, particularly for highly variable traits such as AT and AN.

A preliminary multivariate analysis using General Linear Model (GLM) procedure (SAS Institute Inc., 2014) was performed to identify fixed factors significantly affecting the variability of milking and conformation traits. Interaction effects identified in the preliminary GLM analysis were not explicitly included in the final animal model but were accounted for within combined fixed effects (contemporary groups). The fixed effects included in the linear model were: herd (7 levels), year of calving (2012, 2013, 2014), calving season (summer: March–August; winter: September–February), and stage of lactation at the time of conformation scoring (5–60, 61–90, > 90 days). First-order interactions among these factors were also considered:

Because all these factors were found to be statistically significant, they were included in the linear model used to estimate heritability, as well as the genetic (r _G) and phenotypic (r _P) correlations. In order to estimate the variance and covariance components, the two-trait animal model was given as follows:

The GIBBS1F90 program (Misztal et al., 1992) and a linear two-trait animal model were used. In each analysis, 100,000 samples were generated, of which the first 30,000 were discarded, treating them as the so-called burn-in period. The number of “burn-in period” samples was determined using the POSTGIBBS1F90 program (Misztal et al., 1992).

The genetic parameters of the controlled traits were calculated using the following formulas:

1. Heritability: $h^2=\frac{{\sigma }_a^2}{{\sigma }_a^2+{\sigma }_e^2}$,

2. Genetic correlations: $r_{\mathrm{G}}=\frac{{\sigma }_{a12}}{\sqrt{{\sigma }_{a1}^2+{\sigma }_{a2}^2}}$,

3. Phenotypic correlations: $r_{\mathrm{P}}=\frac{{\sigma }_{p12}}{\sqrt{{\sigma }_{p1}^2+{\sigma }_{p2}^2}}$,

where, σ ² _a – additive genetic variance, σ ² _e – residual variance, σ ² _p – phenotypic variance, σ _a12 – additive covariance between traits x and y, σ ² _a1, σ ² _a2 – additive genetic variance of traits x(1) and y(2), respectively, σ ² _p1, σ ² _p2 – phenotypic variance of traits x(1) and y(2), respectively, and σ _p12 – phenotypic covariance between traits x and y.

3. Results

4. Discussion

5. Conclusion

Traits with moderate to high heritability, such as HS, TL, RUW, RUH, CL, and ME, represent effective selection targets for improving cow adaptability to AMS. Udder conformation significantly influences AMS attachment success, with shallower, well-attached udders and optimal teat placement contributing to improved ME. In addition, balanced body conformation, including appropriate BD, body width, and correct leg posture, enhances both MY and performance in robotic milking systems. Most phenotypic correlations were weak, indicating that selection decisions should rely primarily on genetic evaluation rather than visual assessment alone. The integration of AMS-derived data into dairy cattle breeding programs may accelerate genetic progress toward cows that are optimally suited for robotic milking environments.

Acknowledgments

The authors declare no additional acknowledgments.

Funding information

Study funded under statutory activity, topic BN-WHiBZ-6/2025. Financially supported by the Minister of Science under the program “Regional Initiative of Excellence” (RID/SP/0017/2024/01).

Author contributions

Conceptualization, methodology, D.P. and B.S.; Investigation, D.P. and G.A.I.; Resources, D.P. and B.S.; Data curation, visualization, D.P. and G.A.I.; Writing – original draft, G.A.I., B.S. and D.P.; Supervision, D.P.; All authors have read and agreed to the published version of the manuscript.

Conflict of interest statement

Authors state no conflict of interest.