More work needed to put dairy data gathering to good use, say researchers

Dairy Focus: University of Guelph researchers say cow health events can be predicted with data

The proliferation of artificial intelligence in today’s dairy barns should eventually lead to the ability to predict a daily disease probability for lactating animals, says University of Guelph researcher Meagan King.

However, one of her University of Guelph compatriots Stephen Leblanc noted that “we’re not there yet.”

King, a post-doctorate animal biosciences researcher and Leblanc, a population medicine faculty member, made presentations during the university’s recent annual Dairy Research Symposium.

Why it matters: The introduction of new sensors and monitors in milking systems occurs regularly, but both speakers stressed the verdict is still out on whether it is worth it over the long-term to invest in such technology.

Related Articles

Leblanc said the ability of new products to carry out monitoring effectively is variable. But the bigger challenge is synthesizing all the information.

King agreed and noted that the 10 per cent of Canadian herds being milked by robots will increase. As a result, more and more producers will have access to the data the units are capable of providing — including milking station refusals, progesterone levels in milk as an indicator of pregnancy status, milk yield, heat cycles and rumination.

“It’s important that we have this data put into actionable, useful reports that don’t overwhelm producers,” said King.

Two weeks before a twisted stomach diagnosis, King said that rumination time and milk yield can show the direction the cow is heading.

Milking frequency before a mastitis diagnosis is different from the milking frequency of a cow unaffected by mastitis at a similar stage of lactation.

In both cases, the synthesis of two pieces of sensor-generated information could lead to early diagnosis and treatment.

Leblanc sees these technologies working best when they augment the human rather than replace. Temperature monitors for fresh cows, for example, could identify those at risk of mineral deficiencies, allowing producers to put their focus on those animals at risk. It has also been shown that humans, especially those who work closely with the cows on a daily basis, are slow to recognize progressive conditions such as lameness. Yet picking up the early signs gives producers the best chance at effectively diagnosing and treating.

However, that is not happening in the vast majority of robot barns. Leblanc, the U of G’s Research Program Director for Animal Production, says work is ongoing to bridge that gap.

“We can collect a ton of data from a cow every day,” he told the symposium audience. “Some of it is clearly useful. Some of it we really don’t know how to use, or we don’t know if it will be useful in the future.”

In contrast to King’s graphs, Leblanc showed one depicting daily somatic cell count levels with a few spikes that quickly drop off. Is this a sign that the cow’s immunity system is working effectively, he asked? Or should this cow be treated immediately?

“We don’t really know,” he said.

“Farmers have these data. They have these cool graphs. But we’re not in the position to say what they should do about that,” he said.

“We’re working on it.”

About the author

Contributor

Stew Slater operates a small dairy farm on 150 acres near St. Marys, Ont., and has been writing about rural and agricultural issues since 1999.

Comments

explore

Stories from our other publications