The changing face of livestock research
Artificial intelligence will enable more precise management of individual animals
By Lilian Schaer for Livestock Research Innovation Corporation
As the issues facing livestock producers the globe over are changing in their complexity, so is the research being done to address those issues.
Topics as complex as greenhouse gas emissions, One Health or antimicrobial resistance need a broad spectrum of expertise and new levels of learning and understanding – such as those increasingly becoming possible through artificial intelligence – if practical, workable solutions are to be found.
Enter Jennifer Ellis, assistant professor in Animal Systems Modelling in the University of Guelph’s Department of Animal Bioscience.
She returned to her alma mater – where she’d completed both undergraduate and graduate degrees – two years ago after post-doctoral research at Wageningen University in the Netherlands and working at Trouw Nutrition as a research scientist.
Her particular specialties lie in dairy cattle nutrition modelling and poultry modelling, but she’s also conducted modelling research on swine, beef, veal, and canines. Ellis’ research is focused on using models as a tool to identify patterns in data, increase understanding of how biological systems work to better predict outcomes, and building links between complex nutrition, health, genetics, and management data to help farmers with on-farm decision making.
“We have a big data wave coming at us, but it is worthless if we don’t know what to do with it,” she explains. “We have to harness the tools we do have for better opportunity analysis and decision-making on the farm. “
That’s where artificial intelligence tools, which make up a core part of her modelling work, can help. According to Ellis, artificial intelligence, in its simplest terms, represents an entire knowledge field that is then broken down into various sub-sectors like machine learning, for example – a bit like science is sub-categorized into specific fields like biology or chemistry.
It covers computer systems that can perform tasks that normally require human intelligence, such as speech or image recognition or decision-making.
“Machine learning is a sub-group within artificial intelligence that focuses on the development of algorithms to make predictions based on detecting patterns in data,” she says. “And within machine learning, there are sub-groups that approach modelling outcomes in different ways but in the end, all try to find patterns in available data and correctly predict outcomes.”
For the livestock industry, this means a greater ability to track individual animals instead of simply a pen, a herd, or a flock, and to do so without intensive labour involvement by farmers or farm workers. It opens up new avenues for precision management, nutrition and feeding that can ultimately lead to healthier, more productive animals.
Examples include activity sensors that monitor and analyze behaviour and can alert producers to possible health problems before an animal shows actual clinical signs of illness or image analysis that can count livestock or estimate their body weight.
“Precision feeding is one area where we are already seeing some prototypes come to the surface, mainly for swine and poultry in research environments, where a system will estimate the weight and growth trajectory of an animal and customize its nutrition accordingly,” she says. “Also, if we can use technology to detect a health event before a vet can, that can offer huge economic return to the producer.”
Another area is precision nutrition formulation at the feed mill through the prediction of pellet quality and durability by analyzing the nutrient composition of inputs. Real-time formulation adjustments could help feed be more consistent.
Overall, though, Ellis notes that artificial intelligence applications have seen much slower uptake in the livestock industry than other agriculture sectors, like crop production, for example. She attributes part of that to the sheer complexity of decision-making on-farm when livestock is involved – and the fact that it can be difficult to feel comfortable handing over the reins to a smart system.
“We can predict a lot of things already, but we also need to know what to do with that information. We are still in exploratory days with how we generate value from all this and ultimately, how we can improve efficiency, increase productivity, and make life easier on the farm,” she says.
Ellis currently has two research projects in this field underway. One, in partnership with the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA) and Trouw Nutrition, is using machine learning algorithms to predict pellet quality at the feed mill.
The other, a collaboration with OMAFRA, Trouw and Wageningen University, brings together artificial intelligence and mechanistic modelling to develop a smart precision nutrition system that ties biological knowledge to predictability.
Wi-Fi and data accessibility are still key hurdles for these technologies to overcome if they’re going to see broader acceptance in the livestock industry, according to Ellis, as is the value proposition. To be successful, the return to producers on their investments must be clear.
“Putting the technology on the farm will also require a new kind of knowledge for producers and on-farm consultants to have, and we in academia need to touch up on this as well so we can train the next generation of producers to be comfortable with these technologies,” she adds. “We are just at the beginning of it, but there is a lot of opportunity.”
This article was published in the August 2021 edition of Ontario Dairy Farmer.