Milking robot data making inroads in breeding value estimation

Robot breeding values have been available in the Netherlands and Flanders since 2015. Using data from milking robots, CRV calculates (among other things) a breeding value for robot efficiency. This enables livestock farmers to specifically breed a herd that can be milked using an automatic milking system.

But the milking robot records other data too that are useful for the breeding value estimation. For example, milking robots measure the average milking speed in kilos of milk per minute of milking time. They also record the position of the teats. “The use of that milking robot data in the breeding value estimation has a number of advantages,” says Gerben de Jong. He is the head of the Animal Evaluation Unit (AEU) at Cooperation CRV, which calculates the breeding values in the Netherlands and Flanders. “Data are collected automatically; we get repeated measurements, with faster availability of data,” he summarises.

Data from more than 1,400 robotised farms

As an example, De Jong mentions the way in which AEU now obtains data about milking speed. “Livestock farmers score their heifers for milking speed during a farm classification visit. But the classifier only visits a farm once every eight months. Furthermore, farm classification visits are not made to all farms. Also using milking robot data in the breeding value estimation therefore allows us to also use data from more animals,” he explains. “And that additional information allows us to calculate better breeding values.”

For the breeding value estimation, CRV currently uses data from more than 1,400 farms that use a Lely milking robot and have provided consent through JoinData for their data to be used. “Lely is a little further along with sharing data through JoinData than other milking robot manufacturers,” says breeding values researcher Niek Meijer, explaining why CRV uses Lely data. “The Lely milking robot also records the teat coordinates. This provides additional information for a number of udder traits.”

In April, CRV will process the robot data of more than 300,000 cows to calculate the breeding value milking speed. A third of these data come from cows that do not have a farm classification score for milking speed. Of all milkings, CRV always uses the milking speed of the first milking and of every twentieth milking after that. “If you were to include all daily milkings of a heifer, you would soon have a thousand milkings for each animal. We don’t need that many measurements to ascertain a reliable breeding value,” says Meijer.

This doesn’t mean that CRV doesn’t do anything with the milking speed data collected through farm classification visits. Those data – from a total of 4.4 million cows – are retained. The livestock farmers also continue entering the milking speed during the farm classification visit. “This way, cows without robot data are also given a reliable breeding value for milking speed,” Meijer explains. Henceforth CRV will calculate two breeding values for milking speed: one based on robot data and one based on data from farm classification visits. “But the breeding value based on robot data will be the main trait,” says Meijer.

The genetic correlation between the two breeding values is high at 0.99. This means that the ranking of the bulls remains virtually the same with the new breeding value milking speed. “The method of collecting data changes with the two breeding values, but the result is the same,” says Meijer.

At 0.51, the heritability of the milking speed based on robot data is higher than that of the milking speed based on a farm classification visit (0.23). “Those repeated measurements allow us to better assess which part of the differences in milking speed are genetically determined,” states Meijer. The reliability of the breeding value milking speed also improves. Especially young bulls with few daughters as yet in the farm classification visit benefit greatly from the daughters with repeated observations already included in the robot data.

Udder balance completely new trait

From the robot data, CRV doesn’t only use the milking speed. The teat coordinates, used by the robot to attach the milking cluster, are also suitable for inclusion in the breeding value estimation. CRV obtains the distance in millimeters from each teat to the midline straight through the udder, the distance to the robot arm, and the distance from the teats to the floor (see illustration to the right).

That provides data on udder depth, about the distance between the front teats and between the rear teats, and about the balance between the fore and rear quarters of an udder. “We currently use those first three purely as an addition, as a correlated trait, when calculating the linear traits for udder,” Meijer explains. “The farm classification visit remains the basis for the breeding values for these udder traits. But, for example, cows that don’t have a score for conformation through the farm classification visit, can now be given a breeding value for these udder traits based on robot data.”

It’s a different story for udder balance. Udder balance is a completely new udder trait that is not scored by Dutch and Flemish classifiers, in contrast to their German colleagues. Also, the trait udder balance cannot be deduced from the scores for other udder traits. “The correlation between the current udder traits and udder balance is between 0.24 and 0.53. In other words: based on the current traits, at the present time nothing meaningful can be said about the udder balance,” says Meijer.

Using robot data, as from April, CRV will be able to publish breeding values for udder balance. “That is especially interesting for farmers that use milking robots. During discussions dairy farmers expressed a desire for that trait which was mentioned a couple of years ago. Robot data now allows us to set to work on this,” Gerben de Jong explains.

“Udder balance is the average difference in udder depth between the rear udder and the fore udder,” explains Niek Meijer. The higher the breeding value, the higher the rear quarters relative to the fore quarters. “But you don’t want the rear quarters to be very high relative to the fore quarters, as the robot will then find it difficult to attach,” says Meijer. “Conversely, you don’t want the rear quarters to be deeper than the fore quarters. Udder balance is therefore an optimum trait.”

In recent years, the average difference in udder depth between the rear and fore quarters has gradually increased, Meijer comments. For example, the difference in black-and-white Holstein heifers in 2014 was 18 millimetres, whilst in 2022 that had increased to 23 millimetres.

There is also a difference between breeds. In Holstein heifers, the difference between fore and rear quarters between 2020 and 2022 averaged 21.2 millimetres. The Jerseys followed with a difference of 7.2 millimetres, whilst in Fleckvieh heifers, the difference was 2.3 millimetres. In the MRIJ heifers, the udder balance was even negative: the rear udder was 2.2 millimetres deeper than the fore udder.

The robot data also clearly show that the difference between the fore and rear udder in Holsteins becomes smaller the older the cows become. “That is also evident in later higher lactations, although the decrease seems to level off slightly in later lactations,” says Meijer.

CRV shows the breeding value for udder balance, just like the other traits, on a relative scale, with an average of 100 and a distribution of 4.

In April, not all bulls will be given a breeding value for udder balance based on robot data. Bulls without daughters with robot data – such as genomic bulls, or bulls imported from abroad – are given a breeding value for udder balance based on their genomic breeding values for the other udder traits and parent information, Meijer adds. “We expect that these bulls will have their own genomic breeding value for udder balance in December.”’

Summary

• As from April, CRV will calculate the breeding value milking speed based on robot data.
• As from April, CRV will publish the new udder trait udder balance, the starting point for which are the teat coordinates gathered by the milking robot.
• Henceforth, the base will be adjusted annually instead of once every five years.
• The breeding values for pregnancy rate cows and maiden heifers will form part of the fertility index.
• The breeding values for direct calving ease will be more stable, with a higher level of reliability.

Teat coordinates in breeding values

When attaching the teat cups, a milking robot uses teat coordinates. These specify the distance from the teats relative to the floor, the robot arm and an imaginary midline straight through the udder. As from April, CRV will include these data in the breeding value estimation, to calculate the breeding value udder balance, among other things.

Fertility and direct calving ease change slightly

In April, the indexes fertility and direct calving ease will change slightly. This means that the breeding value pregnancy rate cows and pregnancy rate maiden heifers will be added to the fertility index. “These additions will give the index a clearer picture of the effect of the bull on fertility,” says Gerben de Jong, head of the Animal Evaluation Unit (AEU) department of Cooperation CRV.

AEU includes both the breeding value pregnancy rate for cows and that for maiden heifers in the fertility index, because there isn’t a high correlation between these two breeding values. “The breeding value pregnancy rate cows measures something else than in maiden heifers,” says De Jong. According to him, it is an additional benefit that more attention is therefore paid to the fertility of maiden heifers. “We see a positive genetic trend for fertility in cows, but it remains at the same level for maiden heifers. It is good that more attention is being paid to this.”

Besides the breeding values for pregnancy rate cows and maiden heifers, the breeding value interval between first and last insemination (IFL) and the breeding value interval between calving and first insemination (ICI) form part of the fertility index.

From now on, in the breeding value estimation for direct calving ease, AEU will take inbreeding into account. The heritabilities for birth traits will also be updated and AEU is increasing the number of observations for direct calving ease per class on a farm. “We currently compare observations from ten consecutive calvings. We will soon be comparing the observations of all calvings within a year,” De Jong explains. This is possible, among other things, because the data collection method has been adjusted. “In the past, livestock farmers recorded the direct calving ease of test bull daughters on charts. Now, via management systems, a lot more data are available about direct calving ease. This ensures that the breeding value direct calving ease becomes more stable and slightly more reliable.”

Base adjustment annually instead of once every five years

This has been happening for some time in countries like Canada and France, but henceforth the Netherlands and Flanders will also adjust the base annually. To date, base adjustment took place once every five years. “We can see that by applying genomic selection, the genetic trend has doubled. Adjusting the base once every five years therefore results in significant changes,” says Gerben de Jong, head of the Animal Evaluation Unit department at Cooperation CRV. “By adjusting the base annually from now on, adjustments will be more gradual.”

The latest base adjustment took place in 2020. Since then, the Dutch and Flemish cows have made progress on several fronts, as shown by table 1. Among other things, they improved their breeding values fertility (+1.3 for black-and-white and +1.6 for red-and-white) and foot health (+1.4 for black-and-white, +1.5 for red-and-white).

For conformation, the cows mainly made progress for the udder and feet and legs, both of which are also taken into account in the NVI. CRV also adjusts the base for production. For kilograms of milk, the red-and-white (+331) is progressing slightly faster than the black-and-white (+289). For kilograms of fat and protein, the black-and-white (+26) have an advantage over the red-and-white (+23).

Because of the base adjustment, the breeding values of the bulls are a better indication of the added value compared to the current cow population. A sample calculation based on the bull Weelder Esperanto shows what effect the base adjustment has on the breeding values of a bull (table 2). In December, Esperanto came in at 321 NVI. With the base adjustment, that 321 minus 69 would equal 252 NVI. His milk output inheritance would go from +1598 to +1309, whilst his breeding value fertility would be 104 instead of 105.

At the same time, CRV is also adjusting the base dual purpose. In the current breeding value estimation, cows with 75 percent MRIJ genes form the base. As from April, those are cows with at least 87 percent MRIJ genes. “Among the cows that determined the base, there were still a reasonable number with 25 percent Holstein blood,” explains De Jong. “They had a significant impact on production, because of which MRIJ bulls and MRIJ cows had trouble obtaining a positive production breeding value. We are now adjusting that.”

Table 1 – Base adjustment for black-and-white, red-and-white and dual purpose

trait	milking goal black	milking goal red	dual purpose
NVI	–69	–63	–35
Inet	–82	–76	4
kg milk	–289	–331	13
kg fat	–14	–12	–1
kg protein	–12	–11	1
lifespan	–157	–110	83
fertility	–1.3	–1.6	–0.7
foot health	–1.4	–1.5	0.6
udder health	–1.3	–1.2	–0.9
saved feed costs for maintenance	10	13	3
frame	–0.4	–0.4	0.0
type	–0.9	–1.2	–0.3
udder	–2.5	–2.7	–1.6
feet and legs	–2.0	–1.1	–1.2
total conformation	–2.6	–2.4	–1.3

Table 2 – Breeding values Weelder Esperanto, with and without base adjustment (breeding values December as starting point)

Trait	current	after base adjustment
NVI	321	252
kg milk	+1598	+1309
lifespan	+674	+517
fertility	105	104
total conformation	109	106