Can Artificial Intelligence Improve Dairy Cow Welfare?

Heat stress is a common welfare problem for dairy cows, especially in the summer months. Regardless of the type of farm, dairy cows can become heat-stressed, both inside a barn (especially if not well-ventilated) or on a pasture (especially if there is no shade). With increasing temperatures worldwide due to climate change, it is no doubt that heat stress will continue to be a welfare problem for dairy cows.

One way to assess heat stress is to use biometric technology available on some farms. For example, modern dairy farms have implemented automated milking systems, sometimes referred to as robotic milking or voluntary milking. Here, cows can choose when in a 24-hour cycle they want to be milked, rather than being milked at a set time. This is done with specialized technology that allows entry to the milking unit by identifying cows through specialized neck collars. These machines then gather data on milk parameters and feed intake, which in turn, can measure heat stress since they are correlated with temperature-humidity indices. Combining this system with artificial intelligence (AI), scientists in this paper have proposed a novel method to reduce heat stress in dairy cows.

Based on one Australian dairy farm, data generated from robotic milkers (e.g., milk yield, liveweight, concentrate feed intake) were collected over the span of four years, along with local weather data (e.g., temperature and humidity). There were 312 cows on the farm. They were all genotyped (by their hair samples) to estimate their level of heat tolerance. For instance, an Australian dairy cow with a “breeding value” under 100 is less heat tolerant than the average cow. Using these breeding values, the authors were able to create two sets of algorithms (one based on all of the cows and another based on a smaller subset of cows with a similar heat tolerance) with machine learning to predict milk yield, milk protein, milk fat, and feed intake.

Two main findings emerged. First, temperature-humidity indices were highest in the summer months and lowest in winter months, as expected. Secondly, their algorithms were successful in predicting milk parameters and feed intake. Specifically, there were strong positive correlations between the observed and predicted parameters in both subsets of cows. Using these algorithms, the authors then proposed how to incorporate AI further onto dairy farms. Their idea is that the AI can track current meteorological data and individual cow data to determine if the cow is heat stressed. If the cow meets the threshold of heat stress, then using the robotic milking system, the AI will open a gate to a milking unit that has water sprinklers. If the cow is determined to be cool, then the AI will direct the cow to the regular robotic milker.

Despite the researchers focussing on increasing milk production as their motivation to propose this AI system, their proposal could improve animal welfare by cooling hot cows. However, since cows are typically only milked once a day, if a cow is heat-stressed, then there’s a chance she can become hot again after leaving the milking station. As such, while this proposal can temporarily cool cows, there is much more that should and could be done to ensure that cows have shade structures, drinking water, misters, and fans throughout the farm. Similarly, this AI proposal will only work for modern farms that have robotic milking installed. While these systems are taking off in Scandinavian countries, the Netherlands, and Iceland, they are much rarer elsewhere. 

For animal advocates, the main takeaway message is that heat stress will continue to be a problem with increasing global temperatures, and that it’ll take more than just AI to alleviate this problem. While systems like this certainly present novel ways that animal welfare could be improved with technological interventions, it is a drop of the proverbial bucket of what needs to be done.